Has the pay-to-performance sensitivity
changed since the financial crisis?
A study of the top 33 banks in the US
Olivia van Roijen
10354093
Bachelor thesis BSc Economics and Business Economics and Finance track Thesis supervisor: Razvan Vlahu June 2015
Abstract
Executive compensation is an important part of corporate governance. This paper examines the relationship between firm performance, firm risk-taking and CEO compensation of the top 33 US banks in the time periods 2003 – 2006 and 2007 – 2013. This study shows that the effect on the different components is more important than the effect on the absolute value of CEO compensation. In this study CEO compensation is split into cash, equity and debt compensation. This study finds a positive relationship between CEO total, cash and debt compensation after the financial crisis. Additionally, we find a negative relationship between risk-taking and CEO total, equity and debt compensation.
Statement of Originality
This document is written by Olivia van Roijen who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of Contents 1. Introduction 4 2. Literature review 5 3. Research methodology 11 3.1 Data collection 11 3.2 Variables 12 3.3 Research design 13 3.4 Descriptive statistics 14 4. Empirical results 17
4.1 Bank performance and CEO compensation 17 4.2 Bank risk-taking and CEO compensation 20
4.3 Robustness checks 22 5. Conclusion 24 References 25 Appendix A 26 Appendix B 26 Appendix C 27 Appendix D 27 Appendix E 28 Appendix F 29
1. Introduction
Executive compensation has been a heatedly debated topic throughout time. Recently, the directors of one of the largest banks in the Netherlands, the ABN AMRO, had given themselves a salary increase of €100,000 a year, which caused a public outcry. In the United States (US), JP Morgan’s CEO was announced to be receiving $20 million as compensation, including a $7.4 million cash bonus and equity awards of $11.1 million. The announcement caused great discontent amongst the shareholders, from whom 38.6% voted against the new compensation package. Even though executives receive high compensation across many different sectors, banks and financial institutions are held to different standards. Bank failure can, depending on the size of the bank, lead to negative externalities such as reduced economic growth, which impacts society as a whole. Therefore, as the health of the banks and the economy are inextricably linked, more attention is given to their compensation policies. For example, in 2007 vast amounts of risk-taking by banks had, partially, led to a financial crisis. As a result, many financial institutions had to be bailed out by their government, to minimize the economic damage. Accordingly, the public, whom has been affected by the crisis, does not feel the bank executives deserve to receive enormous amounts of compensation.
The reason firms advocate for executive compensation, such as bonuses and equity awards, stems partly from the corporate governance strategy. The interest of shareholders and the executive management are not aligned, which is known as the agency problem. To mitigate the effects of the agency problem, executive management needs to be incentivized to take actions that enhance shareholder wealth. However, the existing literature has often disagreed on the actual effects of firm performance on executive compensation. Where some empirical studies find significant effects other studies do not. Another reason for high compensation stems from the most basic economic principle of demand and supply. Where there is a high demand and low supply of labor, the price will rise, which is the case in the labor market of top executives.
The research question of this paper is:
To what extent did bank performance and bank risk-taking affect CEO compensation before and after the start of the financial crisis?
The main purpose is to empirically examine the relationship between CEO compensation and bank performance from 2003 to 2013. One of the fundamentals of economics is that people respond to potential rewards, which does not necessarily mean cash.
Studies have shown that the composition of the compensation package is more important than the total compensation. Hence, this paper studies the impact of three different types of compensation: cash, equity, and debt. The time frame is split into two separate time periods to distinguish between pre- and post-crisis data. To narrow down the research, the study looks at the largest 33 banks, with respect to total assets, in the US. European banks are not included as the information is less transparent and the data, therefore, not readily available. For the pre-crisis time period, this study only finds a positive relationship between bank performance and debt compensation. However, for the post-crisis period, a small, but significant, positive relationship between bank performance and total, cash and debt CEO compensation exists. Between bank risk-taking and total and equity compensation a negative significant relationship is also found, confirming the hypotheses of this paper.
The remainder of this paper is organized into six sections. Section 2 gives an introduction to the economic theory, explains the types of compensation and reviews the existing literature. Section 3 elaborates on the research methodology, including the hypotheses, the data, and the method used. Section 4 presents the empirical results and section 5 provides a discussion of the main findings and concludes.
2. Literature review
Corporate governance is a diverse topic in the academic literature. It deals with mechanisms by which stakeholders of a firm exercise control over the firm’s management. The stakeholders include equity- and debt-holders, employees and customers amongst others. The management makes the key decisions and therefore, stakeholders need to find a way to ensure that their interest is protected (John and Senbet, 2998). This separation of ownership and control leads to an agency problem, which is discussed in the following paragraph. Board effectiveness, ownership structure and compensation have been studied repeatedly. The primary way for equity holders to exercise control is through the board of directors. John and Senbet find that board effectiveness is determined by its independence, size and composition. Given the diversity of studies and outcomes, there is ambiguity about the influence of board and ownership structure on executive compensation and firm performance (Core et al., 1999). There have been many studies showing why certain mechanisms of corporate governance fail, but according to Core et al., an optimal governance structure has not been found yet.
There are a few strands of literature related with this paper. First, the concept of agency theory is addressed. An agency relationship is defined as a contract under which one or more principal engages with one or more agent to perform services on their behalf (Jensen and Meckling, 1976). Assuming both parties are utility maximizers, Jensen and Meckling argue that the agent might not always act in the best interest of the principal. The principal can limit the deviations from their interest by awarding appropriate compensation and by monitoring the agent. This leads to three types of costs: monitoring expenses, compensation expenditure, and the residual loss. Residual loss is the loss of the decreased welfare of the principal due to the agent acting in self-interest. Within a firm, shareholders and executives want to maximize their own wealth. This conflict of interest, between shareholders and the executive management, is a typical example of the principal-agent problem. In this example, the shareholders are the principals and the executive management the agents (Jensen and Murphy, 1990b). According to Jensen and Murphy, shareholders want executives to maximize profit, but the executives weigh off their private benefits and costs before taking any actions. In these situations, agency theory suggests that a compensation policy should be implemented to incentivize executives to undertake actions that enhance shareholder wealth (Jensen and Murphy, 1990b).
Another issue related to agency theory is the problem of risk sharing, which arises when the principal and agent have different risk preferences (Eisenhardt, 1989). Executives tend to prefer less risky investments than the shareholders, even if the riskier projects are more profitable (Faleye and Krishnan, 2014). Therefore, shareholders who want to increase profitability might encourage risk-taking. Faleye and Krishnan suggest that in the secondary sector shareholders choose the compensation to encourage risk-taking. For example, by introducing a variable compensation component related to a target or other measure. However, in the financial sector this is not the case as risky behavior by financial institutions is not desirable. Financial intuitions, such as banks, have a great impact on the economy as a whole and risky behavior can lead to great negative externalities. An example of negative externalities is the financial crisis of 2007, which was partially caused by risky behavior by banks. A financial crisis has an effect on society as a whole. Therefore, the compensation policy at financial institutions should be considered carefully. The trade-off between the objectives of providing efficient risk sharing and providing the executives with incentives to take appropriate actions must be reflected in the compensation policy (Jensen and Murphy, 1990b).
A second strand of the literature is concerned with the banking sector. Laeven (2013) identifies several aspects in which banks differ from nonfinancial firms, with leverage being one the principal differences. Banks are highly leveraged, meaning they have a large amount of debt compared to equity. Due to the high amount of debt, shareholders encourage risk-taking as creditors carry most of the downside risk. If the bank goes bankrupt the losses for shareholders are limited whereas the losses for debt holders can be colossal (Laeven, 2013). Secondly, the debt of a bank is generally short term and in the form of many small deposits from several different depositors. These elements increase the liquidity risk of a bank, meaning it is very easy for a bank to transfer its assets into cash (Laeven, 2013). Another difference Laeven mentions is the systematic importance of banks to the economy. The risk that a bank takes is a risk to the entire financial system. In particular the relatively large banks are a risk to the economy, therefore they are sometimes labeled as ‘too big to fail’. This means the bankruptcy of a large bank will have such a great negative impact on the economy as a whole it has to be avoided. In these situations, the government gives out a de facto guarantee, which means the government bails out the bank with government funds if necessary (de Haan and Vlahu, 2013). Consequently, creditors perceive this as a reduced risk, as the government provides the bank with funds if it cannot meet its obligations (de Haan and Vlahu, 2013). As shareholders are protected by limited liability, they would prefer the bank to increase the level of debt and take more risks to increase shareholder wealth. However, this causes great concern for the society as a whole as it increases the systematic risk of the economy (Laeven, 2013). To resolve this issue, strict regulations in the financial sector have been put into place.
A third strand of the literature is compensation. Executive compensation is an important mechanism of corporate governance. According to Jensen and Murphy (1990a), in most publicly held companies, the compensation of top executives is virtually independent of performance. The pay-to-performance sensitivity is defined as the change in the CEO’s wealth for every change in shareholders wealth. Jensen and Murphy (1990b) show the CEO wealth rose by only $3.25 for every $1,000 in firm value. On the other hand, in a more recent study, Hall and Liebman (1997) find a very strong pay-to-performance relationship, especially with respect to equity compensation. They find that compensation and the sensitivity to firm performance had increased substantially from 1982 to 1997. The difference between the two studies can be contributed to the fact that the former focused on a time
period before the sharp increase in stock option issuance. This shows it is important to distinguish between different types of compensation as each can be affected differently by firm performance and risky behavior.
There are three different types of compensation discussed in this paper. The first type of compensation is cash compensation, which is divided into two sub categories:
- Fixed cash compensation: the base salary paid;
- Variable cash compensation: a bonus that depends on an achievement or a performance measure.
Jensen and Murphy (1990a) find that, on average, CEOs receive about half of their cash salary in the form of bonuses. This shows that the size of the bonus has a significant effect on total cash compensation. Jensen and Murphy (1990b) find a low but positive relationship between cash components and firm performance when looking at a sample of corporations in the time period 1974-1986. The findings of Jensen and Murphy agree with a previous study, which finds an elasticity of CEO salary and bonus with respect to firm value of 0.1. Murphy (1998) finds that pay-performance sensitivities have increased severely from the 1970s to the 1990s. He finds a $0.088 change in CEO salary and bonus for each $1,000 change in shareholder wealth in financial firms. Hall and Liebman (1997) find similar results in a 15-year panel data set of CEOs of the largest US firms. They estimate a pay-to-performance elasticity of 0.19, with a significant increase since the early 1980s. According to their study, the link between the bonus component and firm performance is slightly stronger than the link with the salary component.
Equity compensation is compensation in the form of stock and stock options. Jensen and Murphy (1990a) find the most powerful link between shareholder and executive wealth to be direct stock ownership by the CEO. Nevertheless, between 1980 and 1990, this has decreased drastically. According to Jensen and Murphy, the most important variable to look at is shares owned by the CEO as a percentage of total shares outstanding opposed to, for example, the value of the shares. A significant percentage of total shares ensure the CEO is directly affected by a change in the share price. However, according to Jensen and Murphy, 90% of CEOs own less than 1% of company stock, which is not very substantial. Jensen and Murphy (1990a) show that CEO wealth rose by $2.00 for every $1,000 increase in firm value. Murphy concludes that stock-based compensation accounted for most of the variation in executive wealth, which did not imply that CEOs are mainly focused on stock compensation. He believes that CEOs could not understand precisely how their actions affected the stock value, but could grasp the cash compensation easier and, therefore, focused
more on this part of the compensation. Hall and Liebman (1997) find equity compensation to be the strongest determinant in the relationship between compensation and firm performance. Stock gives CEOs direct equity, whereas stock options only provide CEOs with the difference between the current share price and the exercise price, if they exercise the option. Therefore, calculating pay-to-performance sensitivities for options is more complex. It requires the exercise price and the expiration term information of each option (Murphy, 1998). Using the Black-Scholes valuation method of options, a value for the option can then be computed. However, the use of the Black-Scholes valuation method for the pricing op options has been a debated choice, receiving both positive and negative feedback. However, this is beyond the scope of this study. The use of stock options has increased severely over the past decades, Hall and Liebman (1997) argue that this could be because option grants are less visible and concrete than salaries and bonuses, meaning less public opposition to high compensation for CEOs. Additionally, equity compensation increases the pay-to-performance relationship, as CEOs only earn more money if the shareholders do.
According to Jensen and Murphy (1990a) a problem with equity compensation is that there is no contractual constraint that discourages CEOs from selling their shares. Once the shares have been sold the benefits of equity compensation disappear and the characteristics of cash compensation apply. However, Hall and Liebman (1997) state that many companies have restrictions on selling the stock, with guidelines indicating how much stock the CEO must hold.
Lastly, there is debt compensation, which consists of pension payments and deferred compensation. Whereas most compensation policies and studies about compensation focus mainly on equity, a firm consists of equity and debt. The reason of incentive compensation is to increase firm value, therefore, it only makes sense to value compensation according to equity and debt (Edmans and Liu, 2010). From an equity perspective, it is only interesting to know whether a firm goes bankrupt or not. When a firm goes bankrupt, an equity holder rarely gets its initial investment back. A debt holder, on the other hand, is more likely to receive the money he is owed and, therefore, a debt holder is interested in an as high as possible recovery value. Executives influenced by equity-holders and are, therefore, are motivated to keep the company solvent. To do so, they may sacrifice the recovery value to maintain the solvency (Edmans and Liu, 2010). Edmans and Liu argue that executives can affect the liquidation value. Therefore, their compensation package should be chosen to ensure they are motivated to maintain a high liquidation value. In other words, equity compensation ensures a CEO increases current firm value whereas debt compensation
ensures a CEO improves firm value in case of bankruptcy (Habib and Johnsen, 2000). Jensen and Meckling (1976) argue for an equal proportion of equity and debt compensation but Edmans and Liu favor an equity bias, a greater share of compensation in equity than debt. Edmans and Liu conclude that debt is mainly desirable in firms where there is a real chance of bankruptcy. As the debt compensation acts to increase firm liquidation value, it is unnecessary if there is no risk of bankruptcy. In the financial crisis of 2007 a lot of bondholders lost significant amounts of money. This shows that there is still an unsolved agency problem in the area of executives and debt-holders (Edmans and Liu, 2010).
A fourth strand of the literature is the financial crisis. One of the causes of the financial crisis, starting in 2007 in the US, is argued to be the excessive risk-taking by banks and other financial institutions. Which is partially blamed on weak corporate governance (de Haan and Vlahu, 2013). This risk-taking could be attributed to the poor incentive structure of executives. It encouraged risk-taking to enhance short-term performance without considering the long-term effects and negative externalities (Fahlenbrach and Stulz, 2011). According to Edmans and Liu (2010), the large bondholder losses can be contributed to the absence of debt compensation. However, Falenbrach and Stulz find that CEOs of banks who performed poorly in the financial crisis did not do so intentionally, but genuinely thought they were maximizing firm value.
In summary, there have been many studies about the relationship between executive compensation and bank performance. However, there is no well-defined conclusion. Where Jensen and Murphy (1990a) find that firm performance and compensation are virtually independent, Hall and Liebman (1997) argue that there is a small, but significant, relationship. There hasn’t been a lot of research into the effects of debt compensation, especially in the banking sector. Large banks are not likely to go bankrupt due to the ‘too big to fail’ principle. As debt compensation is of greater importance if there is a higher risk of bankruptcy, this type of compensation might not be as effective. Banks whom do not satisfy the ‘too big to fail’ principle could benefit greatly from debt compensation as banks are highly leveraged and, therefore, the effect of bankruptcy and the liquidation value is very important. This paper intends to examine the effect of bank performance and bank risk-taking on compensation to establish if there is a significant relationship between the variables. Previous literature indicates that the composition in different components is more important
than the total value of compensation. Therefore, this study looks at cash, equity and debt compensation separately. There have already been a lot of studies showing the effects of the different types of cash compensation, therefore, this paper aggregates the cash components and analyzes them as one component. As the financial crisis played an important part in the debate about CEO compensation and risk-taking the data is be split into two time period: 2003-2006 and 2007-2013. This allows us to distinguish between the pre- and post-crisis behavior of compensation and its relationship to bank performance and risk-taking.
3. Research methodology
This study is concerned with the relationship between bank performances, bank risk-taking and CEO compensation of banks in the US. To examine these relationships, hypotheses have been put into place. All hypotheses are tested over two different time periods, 2003-2006 and 2007-2013, to distinguish between pre- and post-crisis relationships. The first hypothesis assumes the relationship between bank performance and total CEO compensation to be positive whilst the second hypothesis assumes the relationship between bank risk-taking and CEO compensation to be negative. Both hypotheses are split into three separate hypotheses to test the effect of the individual compensation components.
Hypothesis 1: Bank performance is positively linked to CEO compensation.
Hypothesis 1.1: Bank performance is positively linked to CEO cash compensation. Hypothesis 1.2: Bank performance is positively linked to CEO equity compensation. Hypothesis 1.3: Bank performance is positively linked to CEO debt compensation. Hypothesis 2: Bank risk-taking is negatively linked to CEO compensation.
Hypothesis 2.1: Bank risk-taking is negatively linked to CEO cash compensation. Hypothesis 2.2: Bank risk-taking is negatively linked to CEO equity compensation. Hypothesis 2.3: Bank risk-taking is negatively linked to CEO debt compensation.
3.1 Data collection
The collected data composes an unbalanced panel dataset with annual observations for the period 2003 to 2013 from the top US banks. The data is gathered from Compustat and Datastream. This paper started with the top 50 banks in the US, ranked by value of total assets. The data for 17 banks was not readily available and, therefore, these banks are not
included in the sample. The list of banks used can be found in Appendix A. The sample consists of a total of 353 bank-year observations.
3.2 Variables
Independent variables
For the first hypothesis, return on equity (ROE) is used as the independent variable as a proxy for firm performance. ROE is the ratio of net income to total equity. It is often used as an indicator of firm profitability relative to its total equity and has frequently been used for bank specific studies (Doucoulagos et al., 2007). For the robustness checks, return on assets (ROA) is used as the independent variable. ROA is the ratio of net income to total assets. The correlation of ROA and ROE in this dataset is 0.9662 as shown in Appendix B. Due to the high correlation, ROA serves as an adequate test variable.
For the second hypothesis the independent variable is the change in bank risk-taking as opposed to bank performance. Bank risk-taking is a very important variable as it relates directly to the principal-agent problem of risk-averse behavior of agents. Risk is proxied by the ratio of risk-weighted assets to total assets, which is an indicator of a bank’s risk according to the Basel I Capital Accord (Berger et al., 2014). Risk-weighted assets are bank’s assets weighted according to their risk, determining how large the proportion of ‘risky’ assets is in comparison to total assets. This data is gathered from Datastream.
Dependent variables
Total compensation is a summation of cash, equity and debt compensation. In order to differentiate between the different types of compensation, they are also tested separately to see how each component is affected by bank performance. All dependent variables are tested in logarithm to study the relative sensitivity as opposed to the absolute sensitivity of pay-to-performance.
The cash component is the summation of the salary and bonus that is earned by the CEO during the fiscal year. The base salary is a fixed component agreed upon prior to the fiscal year, whereas the bonus is variable, usually dependent on a predetermined performance measure or budget goal (Murphy, 1998).
The equity component consists of the stock and options awarded. The stock is defined as the shares owned by the executive, excluding options that are or will become exercisable with 60 days. Options are the values of option related awards. Valuation is based upon the value of options that vested during the year.
Debt compensation is the change in pension value and the non-qualified deferred compensation earnings. This composes of above-market earnings from deferred compensation plans and aggregate increases in actual value of defined benefit and actual pension plans during the year.
Control variable
Prior research finds a significant relationship between CEO compensation and bank size (Jensen and Murphy, 1990b). According to Doucouliagos et al., (2007), all studies about executive compensation have included bank size as a control variable and, accordingly, this paper will too. The variable ‘total assets’ is one of the variables often used as an indication of the size of a bank. The logarithm of total assets serves as the control variable in the first hypothesis.
In the second hypothesis, the logarithm of change in total assets is used as the control variable to account for changes in bank size, which could affect risk-taking behavior (Berger, et al., 2014).
For the robustness checks, two control variables are added to verify this study does not leave out any important variables. The first variable is leverage, which is measured by the ratio of total debt to total equity. For banks this is predicted to be relatively high. The second variable is non-performing assets. Non-performing assets are assets that are not accruing any income, possibly due to default by the borrower.
3.3 Research design
The research technique used in this paper is the Ordinary Least Square method; this is a method to estimate the unknown coefficients in a linear regression model. It minimizes the difference between the observed responses in the dataset and the responses predicted by the linear approximation of the data. Smaller differences prove that the model fits the data. It is important that the repressors are exogenous and there is no perfect multicollinearity between the independent variables. As shown in Appendix B, there is no perfect multicollinearity between the independent variables used as none of the variables have a perfect correlation.
To test hypothesis 1 the following regression model is used:
𝑙𝑛(𝐶𝐸𝑂 𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛)!" = 𝛼 + 𝛽!(𝑅𝑂𝐸)!"+ 𝛽!(𝑠𝑖𝑧𝑒)!" + 𝜀!"
Where the logarithm of compensation is the dependent variable, α is the constant, ROE is the independent variable, size is the control variable measured as the logarithm of total assets, βs are the regression coefficients of company i at time t and ε is the error term which captures
idiosyncratic disturbances. β measures the change in CEO compensation per unit change of ROE or size. According to prior research, we expect to find a small, but significant, relationship between bank performance and CEO compensation, with a stronger relationship for equity compensation (Jensen and Murphy, 1990b). The relationship between CEO compensation and bank size is predicted to be positive, the larger the bank the higher the compensation.
To test hypothesis 2, the following regression model is used:
𝑙𝑛(𝐶𝐸𝑂 𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛)!" = 𝛼 + 𝛽!(𝚫𝑟𝑖𝑠𝑘)!"+ 𝛽!(𝚫𝑠𝑖𝑧𝑒)!"+ 𝜀!"
Where the logarithm of compensation is the dependent variable, α is the constant, change in risk is the independent variable, change in size is the control variable measured as the logarithm of total assets, βs are the regression coefficients of company i at time t and ε is the error term which captures idiosyncratic disturbances. β measures the change in CEO compensation per unit change of risk or size. Bank risk-taking and compensation are predicted to have a negative relationship, as risk-taking in banks should be discouraged due to the large negative externalities. Therefore, lower risk-taking should lead to higher compensation, giving a negative relationship between the different variables. The relationship between CEO compensation and bank size is predicted to be positive, as bank size increases compensation is also expected to increase.
In the robustness checks, additional control variables are included in the model, which alters the regression equation. The new regressions to test the hypotheses are:
𝑙𝑛(𝐶𝐸𝑂 𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛)!" = 𝛼 + 𝛽!(𝑅𝑂𝐸)!"+ 𝛽!(𝑠𝑖𝑧𝑒)!"+ 𝛽!(𝑛𝑜𝑛 − 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑖𝑛𝑔 𝑎𝑠𝑠𝑒𝑡𝑠)!" + 𝛽!(𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒)!"+ 𝜀!" 𝑙𝑛(𝐶𝐸𝑂 𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛)!" = 𝛼 + 𝛽!(𝚫𝑟𝑖𝑠𝑘)!"+ 𝛽!(𝚫𝑠𝑖𝑧𝑒)!"+ 𝛽!(𝚫𝑛𝑜𝑛 − 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑖𝑛𝑔 𝑎𝑠𝑠𝑒𝑡𝑠)!" + 𝛽!(𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒)!"+ 𝜀!" 3.4 Descriptive statistics
In the data set from 2003 – 2006, there are 133 bank-year observations from 33 banks, of which 52 observations are missing the information on compensation.
Table 1
Sample summary statistics for 2003 – 2006.
The table shows summary statistics for the key compensation variables. The data are from Compustat Exucomp. Total is the summation of cash, equity and debt. Cash is the summation of salary and bonus, equity is the summation of shares owned and options granted and debt is the summation of change in pension funds and deferred income. Values are reported in thousands of dollars.
Number Mean Std. Dev. Min Max Total 24 4915.85 6661.12 1058.00 35,123.54 Cash 81 1985.71 2222.37 336.25 14,000.00 Equity 24 2764.49 4132.09 100.15 21,077.09 Debt 24 656.93 7683.26 0.00 3026.61
As shown in first row in Table 1, the average CEO total compensation was $4,915,850, the minimum was $1,058,000 and the maximum was $35,123,540. As indicated by Figure 1, the total compensation consisted of 52% cash, 43% equity and 5% debt compensation, showing that debt compensation was not a large component of total compensation.
Fig. 1. Evolution of the composition of the CEO compensation policy. Cash is the summation of salary and
bonus, equity is the summation of shares owned and options granted and debt is the summation of change in pension funds and deferred income.
In the data set from 2007 – 2013, there are 221 bank-year observations from 33 banks, of which 24 observations are missing the information on compensation.
Cash 52% Equity 43% Debt 5% Composition of compensation 2003 - 2006 Cash 29% Equity 52% Debt 19% Composition of compensation 2007 - 2013
Table 2
Sample summary statistics for 2007 – 2013.
The table shows summary statistics for the key compensation variables. The data are from Compustat
Exucomp. Total is the summation of cash, equity and debt. Cash is the summation of salary and bonus, equity is the summation of shares owned and options granted and debt is the summation of change in pension funds and deferred income. Values are reported in thousands of dollars.
Number Mean Std. Dev. Min Max Total 190 4479.93 3929.09 241.64 32,587.03 Cash 197 1288.91 1359.80 161.54 15,500.00 Equity 190 2354.96 3238.07 80.10 27,995.59 Debt 197 829.70 1395.50 -16.28 10,864.93
Table 2 shows the average total compensation of a CEO was $4,479,930, which is a decrease of $435,920 opposed to the previous time period. The minimum and maximum total CEO compensation were also lower than the period before, this could be due to the financial crisis and the decrease in compensation as one of the consequences. As shown in Figure 1, there is a significant change in composition between the pre- and post-crisis period. Where cash used to be 52% of the total compensation, this decreased dramatically to 29%. Equity compensation increased slightly from 43% to 52% and debt compensation gained 14% of the total share.
Table 3
Sample summary statistics.
The table shows summary statistics for key variables. The data are from Compustat Bank annual databases, except for the variable risk-weighted assets, which is from Datastream. ROE is calculated by dividing net income by total equity and ROA is calculated by dividing net income by total assets. All accounting variables are measured in millions of dollars.
2003 - 2006 2007 - 2013
Number Mean Median Number Mean Median
ROE 132 .14 0.14 220 .04 0.07
ROA 132 .01 0.01 220 .00 0.01
Total assets 132 134,916.00 48,000.99 220 257,721.60 68,450.83 Risk-weighted assets 109 119,958.60 47,568.41 198 192,377.30 63,397.50
The average total assets was $134,916 million in the time period before the crisis, whereas the median was $48,000 million showing there is skewed data, which means it is likely the
mean has been affected by extreme outliers. The same principle holds for the risk-weighted assets. Another remarkable difference is the increase in the mean and median of the total assets and risk-weighted assets between the two time periods, the mean of total assets has nearly doubles and the mean of risk-weighted assets has increased by roughly 60%. The mean and the median of ROA has maintained stable at about 0.01, whereas the mean and the median of ROE have decreased from 0.14 to 0.04 and from 0.14 to 0.07 respectively.
4. Empirical results
The results of the regression analysis are presented in this section and a summary of the results can be found in Appendix C.
4.1 Bank performance and CEO compensation
The hypotheses are designed to test the relationship between bank performance and the aggregate of CEO compensation as well as the individual components. The regression is split into two time periods, 2003 – 2006 and 2007 – 2013 and the results are shown in Tables 4 and 5 respectively.
Table 4
CEO total, cash, equity and debt compensation on ROE from 2003 – 2006.
The table shows results from regressions of the CEO total, cash, equity and debt compensation on ROE and the control variable size. ROE is defined as the net income divided by total equity. Size is defined by the logarithm of total assets. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, **, and *, respectively.
Log total compensation Log cash Log equity Log debt Constant 5.828 (1.195)*** 5.108 (0.683)*** 3.804 (1.770)** 2.481 (4.048) ROE -0.583 (2.990) -1.114 (1.936) 0.990 (4.429) 18.644* (10.127) Size 0.217 (0.112)* 0.209 (0.058)*** 0.307 (0.165)* 0.000 (0.378) Number of observations 24 81 24 24 R-squared 0.1580 0.1421 0.1650 0.1511
Table 5
CEO total, cash, equity and debt compensation on ROE from 2007 – 2013.
The table shows results from regressions of the CEO total, cash, equity and debt compensation on ROE and the control variable size. ROE is defined as the net income divided by total equity. Size is defined by the logarithm of total assets. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, **, and *, respectively.
Ln (total compensation) Ln (cash) Ln (equity) Ln (debt) Constant 6.303 (0.406)*** 5.819 (0.323)*** 5.195 (0.636)*** 0.381 (1.681) ROE 1.489 (0.479)*** 1.149 (0.373)*** 1.035 (0.750) 3.886** (1.967) Size 0.155 (0.036)*** 0.094 (0.028)*** 0.171 (0.056)*** 0.353** (0.147) Number of observations 190 197 190 195 R-squared 0.1458 0.1048 0.0622 0.0527
Hypothesis 1: Bank performance is positively linked to CEO compensation.
The first regression estimates the relationship between the logarithm of total compensation and firm performance measured by ROE. In the pre-crisis time period, only the constant and the control variable are statistically significant at 1% and 10% significance level respectively. The results show that the size of the bank and total compensation are positively related, which is consistent with the prediction of the control variable. However, there is no significant relationship with ROE. In the post-crisiss period, the results for the constant, ROE and size are significant, at a 1% significance level. The result supports the hypothesis and shows a positive relationship between bank performance and total compensation. The regression shows a 1 unit increase in ROE leads to a 1.489% increase in total compensation.
Hypothesis 1.1: Bank performance is positively linked to CEO cash compensation.
The second regression looks at the relationship between the logarithm of cash compensation and bank performance. In the pre-crisis period, only the constant and size variables were significant, which means the hypothesis cannot be accepted. Table 5 shows a significant result for all variables, with a positive relationship between ROE and cash compensation. Therefore, the hypothesis can be accepted, at a 1% significance level, and a 1 unit increase in ROE leads to a 1.149% increase in cash compensation. ROE explains 10.48% of the variation in cash compensation.
Hypothesis 1.2: Bank performance is positively linked to CEO equity compensation.
The third regression estimates the relationship between the logarithm of equity compensation and bank performance. In the pre-crisis period, no significant relationship was found, but the constant and control variable were significant at 5% and 10% respectively. The same analysis can be applied to the post-crisis time period, where the hypothesis cannot be accepted, but the constant and control variable are significant at 1%.
Hypothesis 1.3: Bank performance is positively linked to CEO debt compensation.
The last regression shows the relationship between the logarithm of debt compensation and bank performance. In the pre-crisis period, a significant relationship between the logarithm of debt and bank performance is found at 10% significance level. A 1 unit increase in ROE is estimated to increase debt compensation by 18.644%. In the post-crisis period, the control variable size and bank performance are both significant at 5% significance level. The effect is smaller in this period, a 1 unit increase in ROE here leads to a 3.886% increase in debt compensation.
The first hypothesis estimated the effect of bank performance on CEO compensation. The previous literature predicts a small, but significant, positive relationship between firm performance and CEO cash compensation and an even stronger relationship between firm performance and equity compensation (Jensen and Murphy, 1990b; Hall and Liebman, 1997). In the pre-crisis time period, 2003 – 2006, only the relationship between ROE and debt compensation is positive and statistically significant. This means an increase in ROE leads to an increase in debt compensation. However, we cannot draw any conclusions about the relationship between firm performance and the other compensation components. However, in the post-crisis period, 2007 – 2013, the relationship between CEO total and cash compensation is positively related to firm performance, which agrees with the findings of multiple studies such as Jensen and Murphy and Hall and Liebman. There is no significant relationship between equity compensation and firm performance, even though Hall and Liebman found this coefficient to be large and significant. The change in significance between the pre- and post-crisis time period could be contributed to the financial crisis. As the media more closely studied banks, they were more cautious in giving bonuses and equity awards. Although there has been a relative increase in equity compensation in comparison to total compensation between the two periods, the absolute value of equity compensation fell.
4.2 Bank risk-taking and CEO compensation
The second set of hypotheses is designed to test the relationship between bank risk-taking and the aggregate of CEO compensation as well as the individual components. As before, the regression is split into two time periods, 2003 – 2006 and 2007 – 2013 and the results are shown in Tables 4 and 5 respectively.
Table 6
CEO total, cash, equity and debt compensation on risk from 2003 – 2006.
The table shows results from regressions of the CEO total, cash, equity and debt compensation on risk and the control variable change in size. Risk is defined risk weighted assets divided by total assets. Size is defined by the logarithm of total assets. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, **, and *, respectively.
Log total compensation Log cash Log equity Log debt Constant 6.964 (0.865)*** 5.794 (0.470)*** 5.547 (1.266)*** 6.537 (1.986)*** Δ Risk 0.105 (0.274) 0.022 (0.139) -0.202 (0.400) -5.747 (3.417) Δ Size 0.138 (0.098) 0.165 (0.049)*** 0.253 (0.144)* -0.132 (0.234) Number of observations 18 59 18 18 R-squared 0.1257 0.1702 0.1802 0.1627 Table 7
CEO total, cash, equity and debt compensation on risk from 2007 – 2013 The table shows results from regressions of the CEO total, cash, equity and debt compensation on risk and the control variable change in size. Risk is defined risk weighted assets divided by total assets. Size is defined by the logarithm of total assets. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, **, and *, respectively.
Log total compensation Log cash Log equity Log debt Constant 7.664 (0.296)*** 6.866 (0.237)*** 6.537 (0.450)*** 3.428 (1.215)*** Δ Risk -0.119 (0.039)*** -0.043 (0.031) -0.100 (0.059)* 0.133 (0.375) Δ Size 0.078 (0.032)** 0.015 (0.026) 0.102 (0.048)** 0.147 (0.135) Number of observations 126 128 126 126 R-squared 0.1237 0.0192 0.0625 0.0105
Hypothesis 2: Bank risk-taking is negatively linked to CEO compensation.
The first regression measures the relationship between risks, defined as the change in risk weighted assets divided by total assets, and total CEO compensation. According to Table 6, in the pre-crisis period, the only significant result is the constant. However, according to Table 7, in the post-crisis analysis, the constant, independent and control variable are significant. The coefficient of the risk variable is negative. This indicates that for a 1 unit increase in risk, total compensation decreases by 0.119%, which confirms the hypothesis. According to the R-squared measure, the model can explain 12.37% of the variation in total compensation.
Hypothesis 2.1: Bank risk-taking is negatively linked to CEO cash compensation.
The second regression estimates the relationship between the logarithm of cash compensation and risk-taking behavior of banks. Apart from the constant and the control variable, no significant results are found and the hypothesis cannot be accepted in either time period.
Hypothesis 2.2: Bank risk-taking is negatively linked to CEO equity compensation.
The third regression estimates the relationship between the logarithm of equity compensation and risk-taking by banks. In the pre-crisis time period, only the constant and control variable are statistically significant. However, in the post crisis period there is a significant negative relationship between risk and the equity compensation at a 10% significance level. The hypothesis can be accepted and the model explains 6.25% of variation in equity compensation. As risk increases by 1 unit, equity compensation is predicted to decrease by 0.1%.
Hypothesis 2.3: Bank risk-taking is negatively linked to CEO debt compensation.
The last regression shows the relationship between the logarithm of debt compensation and bank risk-taking. There is no significant result of this relationship in either time period meaning the hypothesis cannot be accepted. The only statistically significant relationship in this regression is between the logarithm of debt compensation and the constant. This means that, according to the estimated line, the elasticity of CEO debt compensation is 3.428% if risk and the logarithm of total assets are zero. But this interpretation makes no sense, as there is no debt compensation if there is no bank, which is the case if risk and total assets are zero. Therefore, the intercept is not economically meaningful.
To answer the second part of the research question, the relationship between bank-risk taking and the different components of CEO compensation is tested. There are no significant results in the pre-crisis time period. However, in the post-crisis time period total and equity
compensation are estimated to have a negative relationship with firm risk-taking, which was predicted by the second set of hypothesis. Possibly, due to the financial crisis banks reconsidered how to discourage risk-taking behavior, to avoid a similar crisis situation. There is no prior literature to compare to these results, but theory suggests that a negative relationship is to be expected. A part of the agency problem is executives being more risk-averse than shareholders, and, therefore, executives need to be rewarded accordingly to incentivize risk-taking behavior to enhance performance. Nevertheless, in the banking industry risk-taking behavior should not be encouraged due to the large negative externalities on society and, therefore, a negative relationship between risk-taking and CEO compensation was expected. The cash component does not have a significant relationship with bank risk-taking. The cash component is an aggregate of salary and bonus, where bonus is a variable part. One explanation of the insignificant relationship could be that a bonus related to risk-taking is not very likely as risk-risk-taking is difficult to measure, whereas it will be reflected directly into equity and debt. As the result is insignificant, no true conclusions can be drawn. The lack of significance in the regression with debt compensation was expected as debt compensation has not been studied yet and, therefore, it is not likely to have been implemented in the compensation strategy.
4.3 Robustness checks
Change of independent variable
To check the robustness of the first set of hypotheses a similar regression is completed, but with ROA as the independent variable. ROA and ROE are highly correlated, as shown in Appendix B, which suggests similar results can be expected. The results of the regressions are shown in Appendix D, Tables D1 and D2, with a separate regression for the pre- and post-crisis respectively.
As shown in Appendix D, Table D1, in the pre-crisis period, the constant is significant in all four regressions, which is equal to the results from the regression with ROE as the independent variable; however, the control variable size is not significant anymore in relationship to the logarithm of total compensation and equity compensation. The post-crisis results are more interesting, the same results are significant but with very different coefficients. As shown in Appendix D, Table D2 the coefficient of ROA in the first regression is 12.67, whereas this is 1.489 in the regression with ROE. In relationship to the logarithm of cash compensation, the coefficient of ROA is 10.947 whereas it the coefficient
of ROE is 1.149. This means the total and cash component of compensation are more responsive to changes in ROA than changes in ROE. As bank are highly leveraged the difference between ROA and ROE is larger than normal, which could explain the difference in the coefficients. Similar to ROE, the relationship between the logarithm of equity and debt compensation and ROA is not significant.
Additional control variables
The R-squared is a statistical measure of how close the data are to the fitted regression line. In the models described above, the R-squared is rather low with an average of 11.47%. This means the model can explain 11.47% of the variation in compensation on average. A low R-squared can be due to omitted variable bias. Adding more control variables to a model can increase the precision of the model. However, it is important to ensure there is no perfect multicollinearity in the model. In the previous model, only the variable total assets is used to control for size. To verify this study does not leave out any important variables, exactly the same regressions are ran for both time periods but with two additional control variables: leverage and non-performing assets. As shown in Appendix E, none of the variables are perfectly correlated, which means there is no sign of perfect multicollinearity.
The results of the regressions are presented in Appendix F. There are many similarities, but also some differences between these results and the results presented above. As shown in Table F1, only hypothesis 1.3 can be accepted in the pre-crisis time period. This agrees with the findings in Section 4.1. However, the coefficient in this test is larger than the coefficient in the original regression. According to the new regression, a 1 unit increase in ROE is estimated to lead to a 21.147% increase, opposed to a 18.644% increase, in debt compensation. As shown in Table F2, in the post-crisis time period, hypothesis 1, 1.1 and 1.3 can be accepted, which also agrees with the original results. The coefficients of ROE are slightly higher, meaning ROE has a stronger impact on ROE when including the additional control variables.
The results to the second hypothesis are displayed in Table F3 and F4. Unfortunately, the number of observations has decreased due to lack of data, which makes the results less valid. Similarly to the original test, none of the hypothesis can be accepted in the pre-crisis time period. Post-crisis, the original test accepted hypothesis 2.1 and 2.2, but with the addition of the two control variables, none of the hypothesis can be accepted.
5. Conclusion
The research question of this study is: To what extent did bank performance and bank
risk-taking affect CEO compensation before and after the start of the financial crisis? In order to
answer this question, previous literature was reviewed and an empirical study was completed. Data on 33 US banks was found and analyzed with an OLS regression technique. The results from this study show similarities, but also differences to previous research. This study finds a positive relationship between bank performance and CEO debt compensation in the pre-crisis time period. In the post-crisis time period, a positive relationship between bank performance and total, cash and debt is found as predicted by the hypothesis. This means that an increase in firm performance leads to an increase in compensation.
As risk-taking behavior in banks was one of the causes of the financial crisis, it seems only reasonable that risk-taking behavior should be discouraged in the post-crisis period. According to this study, this is the case as cash and equity compensation are negatively related to risk-taking. There were no significant results in the pre-crisis time period.
One of the limitations of this study is the data. This study only includes listed banks, as not listed banks are not included in the database. This could have led to biased results, as it does not provide the whole picture of the banking industry. Secondly, all the regressions showed a very low R-squared of around 11%, which means the model does not competently explain the variation in compensation. This could be a sign of omitted variable bias, even when including the additional control variables leverage and non-performing assets, the average R-squared was only 20%. Apart from the economic limitations, the study provides no meaning on psychological factors and cannot account for the level of utility a type of compensation brings to a CEO, one CEO might value equity compensation more than cash compensation and vice versa, and act accordingly. However, this undermines the basic agency theory and would indicate there is no optimal compensation level and composition.
In the future, a similar study into European banks could be beneficial, but this is more complicated as the banking industry is less transparent and data on corporate governance and executive compensation is harder to find. Another interesting direction for further research is the field of debt compensation. Even though debt compensation is only interesting if a firm can risk bankruptcy, it can be a very interesting tool to mitigate the effects. Especially since banks are very high leveraged, a high bankruptcy value is favorable for the creditors and depositors.
References
Berger, A. N., Bouwman, C. H., Kick, T., & Schaeck, K. (2014, April). Bank risk taking and liquidity creation following regulatory interventions and capital support.
de Haan, J., & Vlahu, R. (2013, July 15). Corporate governance of banks: A survey. De
Nederlandsche Bank Working Paper No. 386 .
Doucoulagos, H., Haman, J., & Askary, S. (2007). Directors' Renumeration and Performance in Australian Banking. Corporate Governance: An International Review , 15 (6). Edmans, A., & Liu, Q. (2010). Inside Debt. Review of Finance (15), 72-102.
Eisenhardt, K. M. (1989). Agency Theory: An Assessment and Review. Academy of
Management Review , 14 (1), 57-74.
Fahlenbrach, R., & Stulz, R. M. (2011). Bank CEO incentives and the credit crisis. Journal of
Financial Economics , 99, 11-26.
Faleye, O., & Krishnan, K. (2014, March). Risky Lending: Does Bank Corporate Governance Matter? 23rd Australasian Finance and Banking Conference 2010 Paper .
Habib, M. A., & Johnsen, B. D. (2000). The Private Placement of Debt and Outside Equity as an Information Revelation Mechanism. 13 (4), 1017-1055.
Hall, B. J., & Liebman, J. B. (1997). Are CEOs Really Paid Like Bureacrats? National
Bureau of Economic Research.
Jensen, M. C., & Meckling, W. H. (1976). Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure. Journal of Financial Economics , 3 (4), 305-360. Jensen, M. C., & Murphy, K. J. (1990a). CEO Incentives - It's not how much you pay, but
how. Harvard Business Review (3), 138-153.
Jensen, M. C., & Murphy, K. J. (1990b). Performance Pay and Top-Mangement Incentives.
Journal of Political Economy , 98 (2), 225-264.
John, K., & Senbet, L. W. (1998). Corporate governance and board effectiveness. Journal of
Banking & Finance , 22, 371-403.
Laeven, L. (2013). Corporate Governance: What's Special About Banks? Annual Review of
Financial Economics (5), 63-92.
Murphy, K. J. (1998). Executive Pay. Handbook of Labor Economics , 3.
Appendix A
List of top 33 holding companies in the US used in the paper.
Associated Banc-Corp KeyCorp
Bank of America Corp M&T Bank Corp.
Bank of New York Mellon Corp. New York Community Bancorp Inc.
BB&T Corp. Northern Trust Corp.
BBVA Compass Bancshares Inc. People's United Financial Inc.
City National Corp PNC Financial Services Group Inc.
Comerica Inc. Popular Inc.
Cullen/Frost Bankers Inc. Regions Financial Corp.
East West Bancorp Santander Holdings USA Inc.
Fifth Third Bancorp State Street Corp
First Horizon National Corp. SunTrust Banks Inc. First Niagara Financial Group Inc. Synovus Financial Corp
First Republic Bank TD Banknorth Ing
FirstMerit Corp US Bancorp
HSBC North American Holdings Inc. Wells Farfo &Co. Huntington Bancshares inc. Zions Bancorp JPMorgan Chase &Co
Appendix B
The following table shows the correlations between the main variables of this study.
ROE ROA Log total assets Risk weighted assets Log Δ total assets ROE 1.000 ROA 0.9359 1.000
Log total assets 0.0170 -0.0585 1.000 Risk weighted
assets -0.0587 -0.0951 0.8041 1.000
Log Δ total
Appendix C
The following table shows a summary of the results of the hypotheses.
2003 – 2006 2007 - 2013 H1: Bank performance is positively linked to CEO
compensation.
Not significant Yes
H1.1: Bank performance is positively linked to CEO cash compensation.
Not significant Yes
H1.2: Bank performance is positively linked to CEO equity compensation.
Not significant Not significant
H1.3: Bank performance is positively linked to CEO debt compensation.
Yes Yes
H2: Bank risk-taking is negatively linked to CEO compensation.
Not significant Yes
H2.1: Bank risk-taking is negatively linked to CEO cash compensation.
Not significant Not significant
H2.2: Bank risk-taking is negatively linked to CEO equity compensation.
Not significant Yes
H2.3: Bank risk-taking is negatively linked to CEO debt compensation.
Not significant Not significant
Appendix D
The following tables show the regression results for robustness checks with a different independent variable.
Table D1
CEO total, cash, equity and debt compensation on ROA from 2003 – 2006.
The table shows results from regressions of the CEO total, cash, equity and debt compensation on ROA and the control variable size. ROA is defined as the net income divided by total assets. Size is defined by the logarithm of total assets. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, **, and *, respectively.
Log total compensation Log cash Log equity Log debt Constant 5.809 (1.195)*** 5.239 (0.678)*** 3.781 (1.762)** 7.169 (2.898)** ROA 5.231 (27.420) -23.379 (19.727) 20.613 (40.419)* 40.456 (65.445) Size 0.204 (0.110) 0.211 (0.058)*** 0.297 (0.162) -0.165 (0.255) Number of observations 24 81 24 21 R-squared 0.158 0.1537 0.1732 0.0367
Table D2
CEO total, cash, equity and debt compensation on ROA from 2007 – 2013.
The table shows results from regressions of the CEO total, cash, equity and debt compensation on ROA and the control variable size. ROA is defined as the net income divided by total assets. Size is defined by the logarithm of total assets. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, **, and *, respectively.
Log total compensation Log cash Log equity Log debt Constant 6.249 (0.409)*** 5.785 (0.324)*** 5.158 (0.636)*** 4.067 (1.237)*** ROA 12.670 (4.801)*** 10.847 (3.755)*** 7.972 (7.476) 14.479 (18.172) Size 0.160 (0.036)*** 0.097 (0.028)*** 0.175 (0.056)*** 0.142 (0.107) Number of observations 190 197 190 155
Appendix E
The following table shows the correlations between the independent variables of this study and the added control variables in the robustness tests.
Table E1
Correlation between the independent variable ROE and several control variables in time period 2003 – 2006. ROE Log total assets Leverage
Log non-performing assets ROE 1.000
Log total assets 0.1613 1.000
Leverage 0.2890 0.1214 1.000 Log non-performing
assets 0.1501 0.7931 -0.0178 1.000
Table E2
Correlation between the independent variable change in risk and several control variables in time period 2003 – 2006.
Δ risk Log Δ total assets Leverage
Log Δ non performing assets
Δ risk 1.000
Log Δ total assets -0.2119 1.000
Leverage 0.0955 0.1933 1.000 Log Δ non
performing assets -0.1964 0.5550 0.1376 1.000
Table E3
Correlation between the independent variable ROE and several control variables in time period 2007 – 2013. ROE Log total assets Leverage
Log non-performing assets ROE 1.000
Log total assets 0.0724 1.000
Leverage -0.2736 0.2697 1.000 Log non-performing
Table E4
Correlation between the independent variable change in risk and several control variables in time period 2007 – 2013.
Δ risk Log Δ total assets Leverage
Log Δ non performing assets
Δ risk 1.000
Log Δ total assets -0.0891 1.000
Leverage 0.0424 0.4735 1.000 Log Δ non
performing assets 0.0137 0.6033 0.5864 1.000
Appendix F
The following tables show the regression results for robustness checks with extra control variables.
Table F1
CEO total, cash, equity and debt compensation on ROE from 2003 – 2006.
The table shows results from regressions of the CEO total, cash, equity and debt compensation on ROE and the control variables. ROE is defined as the net income divided by total equity. The control variables are the logarithm of assets, the logarithm of non-performing assets and the leverage ratio. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, **, and *, respectively.
Log total compensation Log cash Log equity Log debt Constant 6.342 (1.488)*** 4.810 (0.0.862)*** 3.903 (2.223)* 5.718 (5.015) ROE -0.448 (3.138) -1.537 (2.027) 0.680 (4.688) 21.147 (10.575)* Log total assets 0.238
(0.174) 0.223 (0.099) 0.415 (0.259) -0.374 (0.585) Log non-performing assets -0.030
(0.133) -0.015 (0.862) -0.117 (0.198) 0.358 (0.447) Leverage ratio -0.063 (0.061) 0.081 (0.036) -0.068 (0.091) -0.120 (0.204) Number of observations 24 81 24 24 R-squared 0.2035 0.1496 0.1966 0.2048
Table F2
CEO total, cash, equity and debt compensation on ROE from 2007 – 2013.
The table shows results from regressions of the CEO total, cash, equity and debt compensation on ROE and the control variables. ROE is defined as the net income divided by total equity. The control variables are the logarithm of assets, the logarithm of non-performing assets and the leverage ratio. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, **, and *, respectively.
Ln (total compensation) Ln (cash) Ln (equity) Ln (debt) Constant 5.877 (0.436)*** 5.971 (0.354)*** 4.430 (0.674)*** -0.490 (1.849) ROE 1.983 (0.530)*** 1.547 (0.414)*** 1.982 (0.818) 4.252 (2.166)* Log total assets 0.143
(0.053)*** 0.026 (0.042) 0.143 (0.082)* 0.447 (0.222)** Log non performing assets -0.017
(0.040) 0.068 (0.033)** -0.024 (0.062) -0.126 (0.171) Leverage ratio -0.070 (0.022)*** 0.017 (0.018) 0.129 (0.035)*** 0.061 (0.092) Number of observations 187 197 187 194 R-squared 0.1949 0.1313 0.1348 0.0628 Table F3
CEO total, cash, equity and debt compensation on risk from 2003 – 2006.
The table shows results from regressions of the CEO total, cash, equity and debt compensation on risk and the control variables. Risk is defined risk weighted assets divided by total assets The control variables are the logarithm of the change in assets, the logarithm of the change in non-performing assets and the leverage ratio. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, **, and *, respectively.
Log total compensation Log cash Log equity Log debt Constant 9.802 (1.765)*** 5.600 (0.470)*** 9.908 (2.509)** 5.396 (2.076)** Δ Risk 0.421 (2.293) 0.053 (0.099) 3.434 (3.258) -5.141 (2.696) Log Δ total assets -0.083
(0.150) 0.205 (0.084)** -0.041 (0.213) -0.270 (0.176) Log Δ non performing assets -0.031
(0.164) 0.004 (0.079) -0.094 (0.233) 0.199 (0.193) Leverage ratio -0.094 (0.155) -0.049 (0.068) -0.183 (0.220) 0.228 (0.182) Number of observations 10 24 10 10 R-squared 0.1917 0.2893 0.2532 0.4819
Table F4
CEO total, cash, equity and debt compensation on risk from 2007 – 2013 The table shows results from regressions of the CEO total, cash, equity and debt compensation on risk and the control variables. Risk is defined risk weighted assets divided by total assets. The control variables are the logarithm of the change in assets, the logarithm of the change in non-performing assets and the leverage ratio. Standard errors are reported in parentheses. Statistical significance at the 1%, 5% and 10% level is indicated by ***, **, and *, respectively.
Log total compensation Log cash Log equity Log debt Constant 6.573 (0.488)*** 6.305 (0.403)*** 4.754 (0.672)*** 3.673 (1.873)* Δ Risk 0.051 (0.106) 0.064 (0.087) -0.028 (0.145) -0.107 (0.406) Log Δ total assets 0.029
(0.063) 0.022 (0.052) 0.048 (0.087) -0.171 (0.239) Log Δ non performing assets 0.008
(0.0.079) 0.028 (0.065) 0.097 (0.109) -0.023 (0.304) Leverage ratio 0.135 (0.048)*** 0.013 (0.040) 0.160 (0.066)** 0.266 (0.184) Number of observations 50 51 50 51 R-squared 0.2785 0.0511 0.2954 0.0493
