The effect of the Dodd-Frank Act on the misalignment of incentives between CEOs and shareholders

(1)

Amsterdam Business School

The effect of the Dodd-Frank Act on the misalignment of

incentives between CEOs and shareholders

Name Floris Peter Kooij

Student number 10203354

Program MSc Business Economics

Specialization Finance

Number of ECTS 15

Supervisor Ilko Naaborg

(2)

2 Statement of Originality

This document is written by Student Floris Peter Kooij 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.

(3)

3

Abstract

This thesis researches the effect of the Dodd-Frank Act on the alignment of interest between CEOs and shareholders. Additionally it examines whether or not there is any difference between the non-financials firms and financial firms. Prior research found a negative relation between the equity based compensation before the financial crisis and firm performance during the first years of the crisis. This study includes variables to test the short-term incentive relationship and long-term incentive relationship with firm performances. Firm performances are measured by monthly buy-and-hold returns over the year after the compensation has been earned. As a robustness check this study includes return on assets and return on equity as the additional approach.

The main findings are in line with previous literature. The short-term incentive coefficient is significant and negative, meaning that the more short-term incentive the CEOs earns the worse the next year’s performance will be. In addition to these findings, this study finds no difference between financial and non-financial firms before the Dodd-Frank Act. However, for the long-term incentive coefficient the results are significant. Prior to the Dodd-Frank Act the relation between owning firm equity and firm performance is negative. After the Dodd-Frank Act this relation became positive for the financial firms. The main conclusion from these results is that the implementation of the Dodd-Frank Act resulted in an improvement of alignment between CEOs and shareholders.

Keywords:

(4)

4

3.1. Methodology ... 17 3.2 Data ... 20 3.3 Descriptive statistics ... 21 4. Analysis ... 24 4.1. Empirical Results ... 24 4.2 Robustness check ... 28 5 Conclusion ... 33 5.1 Conclusion ... 33 5.2 Limitations ... 34 5.3 Further research ... 35 Appendix ... 36 Appendix (A) ... 36 Appendix (B) ... 37 Appendix (C) ... 37 Appendix (D) ... 37 References ... 38

(5)

5

Introduction

1.1 Background information

The recent credit crisis has caused economics and policy makers to review the economic sector in total. The financial sector, to be more specific the bankers, are the primary cause of the recent crisis. A 2013 article from the economist states: “…. it is clear the crisis had multiple causes. The most

obvious is the financiers themselves—especially the irrationally exuberant Anglo-Saxon sort, who claimed to have found a way to banish risk when in fact they had simply lost track of it”. 1

Governments from all over the world started to take actions against the agents of the financial sector. New regulations were implemented to prevent a crisis of this scale in the future. For the American government it was clear that actions had to be taken to reduce the chance of another crisis. On July 21 2010 the “Dodd-Frank Wall Street Reform and Consumer Protection Act” was signed by the Obama Administration. One of the points of this act is to monitor the financial institutions better, with the idea to prevent a new recession, which led to increasing regulation and monitoring on banks.

Regarding executive compensation, the Dodd-Frank Act (2010) included more protection for shareholders. Votes from the shareholders about the compensation package are now required at least every 3 years and additional information about compensation matters have to be disclosed. The Dodd-Frank Act was one of the largest reformats in terms of regulation for the financial sector. Another legislation adopted by the SEC is the requirement for firms to disclose the ratio of

compensation of their CEO to the median compensation of its employees. It is no surprise that not only the financials firms were being tackled by this legislation, but that it also included laws with respect to executives. According to an article from The Telegraph from 2008, it is the corporate governance at banks that encouraged short-term thinking for their executives which led to a lack of accountability between management and shareholders. 2_{The purpose of corporate governance and}

especially executive compensation is to align interest of CEOs with the interest of their shareholders. If the executives were compensated by their firm performances, one would not expect a crisis like this to happen, since it directly negatively influence their wealth fare.

1.2 Research question

This question was also asked by Fahlenbrach & Stulz (2011), where they examine whether or not better aligned CEOs performed better in the early stages of the crisis. Where Fahlenbrach & Stulz (2010) only focus on the year before the crisis (2006) and the performance in the crisis (2007), this study examines beyond that point and takes into account the new regulation implemented in 2010.

1_{See The Economist, “The origins of the financial crisis : Crash course”, 7 September 2013} 2_{See The Telegraph, “Credit crisis caused by poor bank management, says ACCA”, 6 October 2008}

(6)

6 The question asked in this study is whether or not the Dodd-Frank Act improved the alignment of interest between CEOs and shareholders. To be more specific, the full research question is:

“Did the Dodd-Frank Act solve the misalignment of incentives between CEOs and shareholders and is this effect different for the financial sector?”

The research questions provides us with two insights into the problem. First, the way the model and regressions are setup, we can test whether or not there actually is a misalignment

between CEOs and shareholders. What we know from the study of Fahlenbrach & Stulz (2011) is that CEOs that had their interest more aligned with the shareholders, meaning that their compensation was more based on stock returns, actually performed worse in the first period of the crisis. The setup of this study is to test whether we find the same results and whether the results are different

between industries before the implementation of the Dodd-Frank Act. Secondly, this study brings us better understanding of the effects of the Dodd-Frank Act to both non-financial and the financial sector. The distinguish between the two sector is made due to the fact that regulation within the Dodd-Frank Act is mainly focused on the financial sector and only partly focused on the economy in total. Because of the difference in regulation for both sectors, one could argue different outcomes from the regressions.

1.3 Contribution and outline

As stated before, this study is based on the study of Fahlenbrach & Stulz (2011). The main two contributions of this paper are the insights between different sectors, non-financial versus financial, and the insights into the effects of the Dodd-Frank Act after implementation. The conclusion drawn from the previous paper is that CEOs and shareholders interest were aligned, but the expected effect was the opposite. This study also finds this conclusion for non-financial firms before the Dodd-Frank Act and cannot conclude that this effect is different for firms within the financial sector. These results are mainly obtained from the short-term incentive coefficient in the model. When we look at the long-term incentive coefficient, we find that before the Dodd-Frank Act financial firms had a negative relationship between equity compensation and firm performances. Additionally, this effects changed to a positive relation in the period after the Dodd-Frank Act, concluding that the Dodd-Frank Act improved the alignment between CEOs and shareholders.

Section 2 discusses the existing literature regarding the topic of executive compensation. It starts with explaining the theory behind the current problem with compensation packages, then reviews the current papers with respect to compensation versus firm performance. Next, section 2 looks at the content of the Dodd-Frank Act and discusses the important legislations that are

(7)

7 implemented in 2010. After that there will be a brief summary of the literature review section and a hypothesis about what we expect to come out of the regressions based on previous literature.

Section 3 starts with discussing the model currently used for this study and reviews the background of this model based on the paper of Fahlenbrach & Stulz (2011). Next, the data selection will be explained and contains information about where the data is retrieved from, what time period is used for this study and how much of the data is left after the selection. Lastly this section provides descriptive statistics tables about the main variables used in the model. Additional information on the descriptive statistics can be found in Appendix (A).

Section 4 is where the analysis of this study is being discussed. It provides the regression results done in this study and explains the statistical and economical significance of these results. Additionally to the results from the current regression model, section 4 provides results from two robustness tests used to see the difference in effect and whether or not the conclusions drawn from the first model are still valid.

Lastly, section 5 concludes and discusses the limitations of this study. As stated before, the conclusions for the short-term incentive coefficients are the same found by Fahlenbrach & Stulz (2011) before the Dodd-Frank Act and only validate what was already known from their study. Additionally this study provides insight on the long-term incentive coefficient, which went from a negative relationship between equity compensation and firm performance towards a positive relation after the Dodd-Frank Act for the financial sector. It concludes that the Dodd-Frank Act indeed improved the executive compensation alignment for the financial sector. Finally, this section discusses the limitations and suggestions for further research about this topic.

(8)

8

2 Literature review

This section discusses the old and current theories on the topic of executive compensation. Since this subject is widely investigated by numerous academics, this section will only focus on a few main theories that are important for explaining why this study uses the methodology of Fahlenbrach & Stulz (2011). Section 2.1 discusses the the Principal-Agent Theory, where incentives between executives and shareholders are not aligned. Section 2.2 reviews why compensation is used for solving the Principal-Agent problem and how it is implemented. Section 2.3 goes in-depth on the subject of executive compensation and summarizes the current research and theories with respect to the subject. Mainly the paper of Fahlenbrach & Stulz (2011) stands central in this section, as this paper is the foundation of the methodology used in this paper. Section 2.4 discusses the new Dodd-Frank Act and how (de)regulation has affected executive compensation in the past. At last, section 2.5 concludes the papers on executive compensation so far and provides the hypothesis for this study.

2.1 The Principal-Agent Theory

When we discuss alignment of interest between CEOs and shareholders, we often mention the basic Principal-Agent theory. This theory, which is covered by many academics, discusses the general idea of how the ‘Agent’, the executives, and the ‘Principal’, the shareholders, interact with each other in order to maximize utility for both. This section describes how this theory can be exerted in the misalignment problem between executive and shareholder.

One of the first papers that explains the Principal-Agent problem in-depth is from Jensen & Meckling (1976). They describe the agency relation as a contract where one person (the principal) engages with another person (the agent) to complete a service on their behalf which involves assigning some decision making right to the agent. When both parties try to maximize their own utility there is reason to believe the agent will not act in the best interest of the principal (Eisenhardt, 1989). To reduce this risk, the principal creates incentives for the agent to align the interest. They argue that monitoring the agent is an option to reduce this risk, but it comes at a high costs for the principal. Though these solutions reduce the risk, they state that it is impossible for the principal to fully reduce the risk of deviation by the agent. The relationship between shareholders and managers of the firm is a schoolbook example of a pure Principal-Agent relationship.

(9)

9 Jensen (1993) discusses four possible ways to mitigate the agency problem that arises

between executives and shareholders. The first way is to install a board of directors, which monitor, hire, fire and set the compensation of the executives. This relates to the suggestion done by Jensen & Meckling (1976) and Eisenhardt (1989). A big problem with the board of directors is that is has been controlled by current executives. Since the 90s the amount of outside directors on the board is 80% and in case of half of the firms only the CEO remained inside the board (Horstmeyer, 2011). A CEO monitoring himself will lead to inefficient monitoring. Secondly, the capital markets provide

mitigation of the agency problem by forcing firms to make the CEO wealth more linked to stock-price performance. However, capital markets also contribute to the problem by pressuring executives to meet analyst and market expectations. The third option, is the use of the political and regulatory system. The implementations of laws regarding theft or fraud can mitigate the agency problem. Such laws restrict insider trading, price fixing and earnings management. Additionally, regulations have been designed to protect shareholders, such as the latest Dodd-Frank Act. Lastly, the product markets play an important role in the mitigation process. Jensen (1993) argues that in theory, due to the high level of competition, CEOs must exert high effort in order to maintain their position. Value-destroying managers will therefore automatically be punished by the market and by their

shareholders.

Finally, Garen (1994) analyzes a basic principal-agent model to explain variation in CEO salaries and incentive pay. CEO compensation is structured in a way that it trades off incentives with the insurance to act in interest of the shareholders. His model is derived from the “Pay-performance sensitivity” model of Jensen (1990a). Garen (1994) includes a beta calculating, where the monthly stock returns are regressed on the monthly market returns. This beta is used as the relative

performance variable. Although the coefficients found are positive, the statistical significance is low due to the low T-values provided. Garen (1994) concludes that there is little support for the use of compensation based on relative performance.

Contrary to previous papers, Murphy (2012) states that while currently the excessive risk taking of executives is being discussed, historically the challenge has been to stimulate executives to take enough risk, not too much risk. He argues that the average shareholders is well diversified and have smaller stakes over multiple companies, the executives have relatively large stakes in their company and are less diversified, meaning they will naturally be more conservative with respect to taking risk.

(10)

10

2.2 Executive Compensation

The study of Murphy (2012) summarizes the current state of executive compensation. CEO incentives have always been entrenched with the agency theory. Researchers use the ratio of the total equity compensation to total compensation as a measurement of the incentives. Additionally, Murphy (2012) states that the strongest direct link between the CEO’s wealth and the shareholders wealth is through the CEO’s portfolio of stocks. The compensation of the CEO is also most likely to be indirectly linked to performances based on account measurements such as return on assets.

Bonuses might be the best known example when one thinks about executive compensation. From a principal-agent theory perspective, bonuses are a tool to achieve alignment between

executives and shareholders. Nevertheless, bonus packages bring problems with it in reality, meaning that increasing the value of their bonuses does not always lead to increasing the value of the firm. The problem is the design of the usual executive bonus plan. As described by Murphy (2012), the bonus does not kick in before it reaches a certain performance hurdle. After that, the bonus increases linear with the performance and is most likely to be capped at a certain threshold. This design flaw can have negative effect for firm value. First, a CEO that has reached the upper threshold of his/her performance based measurement has no incentive left to increase firm value even further. Second, a CEO that is below the performance hurdle has two options. The CEO stops putting effort into the firm, because it is impossible for him to reach the hurdle, or the CEO takes on too high risk projects, in hope that the projects will pay out and the hurdle will be reached. Healy (1985) was one of the first that discovered that executives use discretionary accrual charges to move earnings to other periods whenever they reached their upper threshold for the current period. As stated before, the CEO’s share ownership is the most direct link to align interest with the

shareholders. The CEO’s percentage ownership of total shares measures how much the CEO gains from a dollar increase in firm value. Calculating the ownership is commonly done by taking the restricted and unrestricted shares of the CEO and divide them by the total shares outstanding.

When taking stock options into account, evaluating executive compensation becomes more difficult. Unlike shares, different options do not share the same characteristics when it comes to portfolio management for the CEO. Options that are far out-of-the-money provide little incentive to the CEO to increase firm value, whereas options in-the-money or at-the-money provide incentives similar to basic shares. Murphy (2012) uses the delta of the option to give them a “weight” to add them to the portfolio. Not all incentives of the CEO are monetary based. For example, the chance of being fired is linked towards the share price. If the stock is performing poorly, the chances of being fired increases. Hence, the CEO has an incentive to maintain a good stock-price in order to keep the

(11)

11 job. This relates to the point made by Jensen (1993), where he argues that competition within the market creates incentive for the CEO to not perform poorly.

2.3 “Compensation vs performance”

In theory, the relation between compensation and performance can be derived from the Principal-Agent problem. The more the compensation is based on firm performance, the stronger the alignment between executives and shareholders will be. Additionally, a better alignment should result in better long-term firm performance. There are numerous papers researching the practical relation between these two.

In the research of Mehran (1995), the point of discussion is whether the form of

compensation causes difference in the performance of the firm. The focus is on the structure of the compensation instead of the level of compensation. He finds that firms with more outside directors tend to pay their executives more equity based compensation. Firm performance is measured by both Tobin’s Q and the return on assets. An ordinary least-squares analysis is used to test the relation between compensations structure and firm performances. The data includes 153 randomly selected firms over the time period 1973 to 1983. Mehran (1995) concludes that there is a positively relation between the percentage of compensation that is equity based and the firm’s performances and a positive relation between the equity owned by the managers and firm performances.

The commonly known way of examining the relationship between executive compensation and firm performances is the approach from Jensen & Murphy (1990b), where they refer to this relation as the “pay-performance sensitivity”. The sensitivity is defined as the value change in the wealth of the CEO corresponding with a dollar change in firm value. This model indicates how much the wealth/compensation of the CEO is determined by the performance of the firm. This relation found by Jensen and Murphy (1990b) nearly tripled in 2003 from the time period of 1974-1986. By the end of 2011 this sensitivity dropped down and was somewhat higher than the levels from 1992 (Murphy, 2012). Additionally, they find that a $1,000 change in firm value leads to a short-term change in CEO salary and bonus compensation of less than 10 dollar cents. The long-term effect for this is 45 dollar cents.

Jensen & Murphy (1990a) research the relationship between CEO compensation with firm performance. For every $1,000 increase in firm value, they find an increase of $3.25 of the CEO’s wealth. The CEO’s wealth is stated as the sum of base salary, bonuses, stock ownership and stock options. Jensen & Murphy (1990a) obtained observations from 1974 to 1986, which includes a total of 2,213 CEOs over all the years in total. They find that CEOs of large firms tend to have less stock and less compensation based on performance measurements than CEOs of smaller firms. Where Jensen

(12)

12 & Murphy (1990a) look at the dollar values, Hall and Liebman (1998) argue that instead of looking at dollar value of the CEO compensation, we should look at the change in CEO wealth for a 1% change in the value of the firm. They used a panel data set of the CEOs in the largest firms from the United States over a time period of 15 years.

A recent approach to this relationship, the foundation paper for this study, is from

Fahlenbrach & Stulz (2011). They examined whether the bad performances of banks during the start of the crisis (2007 and 2008) is related to the CEO incentives before the crisis (2006). The main difference between the approach used by Jensen & Murphy (1990b) and Fahlenbrach & Stulz (2011) is the shift between dependent and independent variable. Their data consists of 77 commercial and investment banks, where they excluded investment advice firms and pure brokers. The average base salary found in their database is 761.5 thousand dollars and a cash bonus of 2,137.7 thousand dollars. They regress the compensation of the executive before the crisis on the performance of the firm in the crisis. They use this approach to test whether executives that were better aligned with shareholders performed better during the crisis. The independent variable for compensation is split between fixed salary, cash bonus, stock bonus and option bonus. They use Buy-and-hold returns to calculate the performances of the banks. Because of the uncertainty about nationalization of banks after 2008, their research only includes July 2007 until December 2008. They conclude that CEO’s that were better aligned with the shareholders actually performed worse during the crisis, which is the opposite of what was expected. The general idea was that bank CEOs knew what was coming with respect to the crisis. The conclusion from their research is that they probably did not see the crisis coming, since most of the CEOs in the banking sector suffered great losses when the crisis started.

In addition to the papers that research the pay-performance relationship, several papers investigate other variables that might influence the relation between executive and shareholders. Jensen & Murphy (1990b) find that the coefficient of firm returns regressed on compensation varies across firms. Barro & Barro (1990) find that compensation does not only depend on stock returns but also on firm size. They study large commercial banks over the period 1982 to 1987. The database contains a sample of 83 banks. Their study is based on Rosen’s (1982) model, which matches CEO compensation and skill with firm size. The compensation of the CEO is assumed to be fully

determined by the quality of the CEO. They start their analyses by examining newly hired CEOs and study how their compensation changes based on stock returns. Using log transformation for both compensation and total assets, they find that for every 1.0 percent increase in assets the

compensation increases with 0.3 percent. Kaplan & Rauh (2010) discover in their research that in the banking industry, individuals get paid more on average. They compare the compensation for

(13)

13 executives in year 1994 with the compensation in 2004. While all individuals earned more over this period, the individuals from Wall Street increases significantly more. They compare the TDC1 and TDC2 total compensation variables provided by Execucomp and find that executives in financial firms have a higher mean total compensation ($4.24M versus $2.77M) and a higher median ($1.51M versus $1.16M). Finally, Adams & Mehran (2003) examine the corporate governance of bank holding companies. Their sample consists of 35 bank holding companies over the period 1986 to 1996. They find that the ratio between stock options for CEOs and base salary plus bonuses is smaller for bank holding companies. Additionally they find that percentage stock hold by the CEOs of these bank holding companies is also lower compared to manufacturing firms. They conclude from these observations that governance structures and compensation structures differ between industries. Kabir & Duffheus (2008) research in their paper the relation between executive pay and

performances. They hand collected data from Dutch listed companies on the Euronext Amsterdam in the period 1998 to 2001. Where firms from the United States are required to disclose their data about executive compensation, Dutch firms are not. The total database consists of 135 firms that disclosed executive remuneration. They regress firm performance on executive compensation and control for industry characteristics, firm characteristics and time. They do not find a positive relation between pay and performance. The independent variable, return on assets, has a statistical negative coefficient of -0,484, meaning that for every 1,0 percent increase in return on assets the executive cash compensation decreases with -0,484 percent. This is consistent with their view that powerful CEOs can influence their own compensation and usually overpay themselves.

Several papers discuss the influence of corporate governance on executive compensation. Brick, Palmon & Wald (2006) research the high compensation for both CEOs and for directors. They find that the firm performance is negatively correlated with the compensation of the CEOs and directors and conclude cronyism in boards. Cronyism is described as the unfair practice by powerful persons, such as politicians or executives, to give jobs and other favors to their friends. The

compensation is divided into cash and total, where total includes all bonuses such as stock granted and stock options. The paper of Core, Holthausen & Larcker (1999) find that CEO compensation is greater when the governance structure is less effective. They find a statistically significant negative relation between compensation and subsequent firm performances. The compensation is the sum of base salary, cash bonus and valuation of the stock options. They value stock options at 25% of their exercise price, saying that the paper of Lambert et al. (1991) suggest that more complicated pricing models find the same results on average.

Finally, items on the balance sheet have been examined whether they influence the effect between performance and compensation. Jensen and Meckling (1976) state that this leads to a

(14)

14 conflict of interest between shareholders and debtholders. Because debt plays an important role in risk taking by the executive, several researches included leverage ratios to measure risk-taking incentives. Murphy (2012) notes that debt itself does not automatically cause more risk taking, but that the limited liability feature of equity leads to this problem. Lastly, Bizjak, Brickley & Coles (1993) use the firm’s R&D intensity as explanatory variable in their regression. They say that R&D

investment should be related with a lower salary link to firm performance. Firms that do not report R&D expenditures, mainly financial companies, are set to zero in their database.

2.4 Regulation – Dodd Frank Act

After the recent financial crisis, executive compensations have become a point of discussion again. The public opinion regarding high executive remunerations is averse, especially towards the financial sector. Regulating high levels of CEO compensation has been an important topic in United States over the past century. There has been some major regulations in the past, for example the reform of high salaries for executives in 1932, the “restricted stock option plans” in 1960s that caused to prohibit option repricing and changing expiration terms, and in the 1990s there was a regulation for non-performance based compensation of more than one billion dollars which could not be deducted anymore as business expense.

The Dodd-Frank Reform Act is in line with previous regulation and continues to manage the abuse in top-level CEO compensation. On 21st_{of July 2010 the Dodd-Frank Wall street Reform and}

Consumer Protection Act was signed by President Obama. This section discusses two major points regarding CEO compensation. The first involves compensation and governance reforms for all publicly traded companies. The second only involves the financial sector, where additional regulations have been put into place to restrict executive compensation.

While the main focus of the Dodd-Frank Act is on regulating firms in the financial sector, some new legislations have been added to improve corporate governance for all large publicly traded firms. Murphy (2012) points out five categories that (in) directly affect the corporate governance on executive compensation for these firms.

1. Section 951 : Say-on-Pay

Every three years, shareholders are asked to approve the CEO compensation plan with an additional vote the first years and every six years to determine when the Say-on-Pay voting takes place (every one, two, or three years). Even though the votes are non-binding, the Say-on-Pay meetings will give shareholders additional control over the executives. The increase in involvement should lead to better alignment between agent (CEO) and the principle (shareholder). Additionally, shareholders are also asked to approve any golden parachute agreements.

(15)

15 2. Section 954: Clawbacks

Firms have to report policies for taking back executive compensation based on financial statements that are afterwards restated. This applies to all payments made in the three years preceding the restatement. The rule applies for both current and former executive member of the firm. This means that when the previous year’s financial statements are restated, the firm can ask the compensation from the executive which was based on the old financial statement and is not in line with the new one.

3. Section 952: Compensation Committee Independence

To improve independency of the compensation committee, the committee must only contain outside independent directors. Independent directors means that there will be check whether or not the director has any financial ties with the firm. Additionally, the firm must check the independence of their compensation consultants, accountants and other advisors to the committee.

4. Section 971: Proxy Access

This section gives the SEC the authority to create new rules allowing certain shareholders to

nominate candidates in the annual proxy statement. According to Murphy (2012), this is potentially the most important of the five categories named. It gives shareholders more rights and more possibilities to replace poor directors with better ones.

5. Section 953, 955, 972 : Additional Disclosures

The additional disclosures concern mostly reporting information about CEO compensation, realized compensation and compensation with respect to firm performances. For example, firms must report the ratio of CEO compensation with respect to the median of all other employees. It is important to note that this information is only being disclosed. The SEC had not implemented rules regarding this, for example, what the consequences are when the ratio of CEO compensation with respect to the median of all other firm employees is too high.

Government involvement has been a response to executive compensation, but also a major driver of it. Murphy (2012) argues that political factors should be taken into account when

researching executive compensation. Most of the literature about executive compensation focus on the relation between the shareholders and the CEO, such as Jensen & Murphy (1990b)

“pay-performance sensitivity”. Habbard & Palia (1995) examine in their paper the CEO compensation in the banking sector and the effect of deregulation. Their main findings are the pay-performance relation in deregulated market is stronger and the CEO turnover rate increases after deregulation. Cunat & Guadalupe (2009) study the effect of deregulation on total payment to the executives and on the pay and performance relationship. A difference-in-difference analysis is used for the banking

(16)

16 and the financial sector. They find that the deregulation substantially changed the compensation of the executives. The fixed compensation decreased over the period and the variable component increased. Both papers find a change in CEO compensation and the structure of it after

deregulations. These changes are important to keep in mind, since deregulation leads to a change, one could argue that regulation will lead to a change in terms of compensation structure. It is therefore important to take regulation and government involvement into account when researching the effects of compensation on performance.

2.5 Conclusion & Hypothesis

The hypothesis of my research states that the coefficients in row three, which represent the relation between executive compensation and firm performance for the financial firms before the Dodd-Frank Act, will not be significantly different from zero. There is no theory that explains why financial firms have a different pay-performance relation than non-financial firms. Barro & Barro (1990) find that the compensation packages differ between industries. When taking this into account, one could argue that there might also be a different effect in the relation between compensation and

performance. Secondly, the main principle of the Dodd-Frank Act was to regulate the financial sector and ensure that CEO incentives are more aligned with their shareholders. Given this goal by the Obama Administration, one would expect that the coefficients for the financial sector will be significant and positive. The coefficients for the financial firms are displayed in row two, before the Dodd-Frank Act, and in row four, after the Dodd-Frank Act. However, previous research suggest that deregulations leads to greater pay-for-performance sensitivity of CEOs at banks (Habbard & Palia, 1995), which could indicate that regulation will cause the opposite effect of this.

Several important conclusion can be drawn from previous literature regarding the research question of this thesis. First, there is enough evidence that supports parts of the Principal-Agent theory. The direct link between an equity stake in the company and firm performance has been proven to be a strong one. The “Pay-Performance Sensitivity” of Jensen & Murphy (1990b) nearly tripled from 1986 to 2003, meaning equity based and performance measurement based

compensation became much popular. Although this suggest a strong link between the two, Jensen & Murphy (1990a) argue that the sensitivity is on average too low to be consistent with the Principal-Agent theory. The structure of the compensation is different for firms in the banking industry, suggesting the relation might also be different for the banking industry. Finally, murphy explains government involvement and regulations do have a major impact on executive compensation and should be taken into account when examining.

(17)

17

3 Methodology and Data

This section will cover the methodology of the model that is used in this paper. The full regression model will be written out and this section will explain how all the coefficients and the entire model will be tested. Lastly, I will discuss how the regression model of Fahlenbrach & Stulz (2011) is build up and the link between the new model and the old. Section 3.2 describes the selection process for the data used in this study. Section 3.3 presents the main descriptive statistics on the data used.

3.1. Methodology

The regression below shows a full version of the new model:

𝑆𝑡𝑜𝑐𝑘𝑅𝑒𝑡𝑢𝑟𝑛𝑖𝑡+1= 𝛾1∗

(𝐶𝑎𝑠ℎ 𝑏𝑜𝑛𝑢𝑠)

(𝐵𝑎𝑠𝑒 𝑆𝑎𝑙𝑎𝑟𝑦)_𝑖𝑡+ 𝛾2∗

(𝑂𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝 𝐶𝐸𝑂)

(𝑇𝑜𝑡𝑎𝑙 𝐸𝑞𝑢𝑖𝑡𝑦) _𝑖,𝑡+ 𝛾3∗(𝐷𝑜𝑙𝑙𝑎𝑟 𝑔𝑎𝑖𝑛 1%)𝑖𝑡 (1)

The first part of this regressions model presents the coefficients for non-financial firms before the Dodd-Frank Act. The next lines of coefficients are an addition to the first regression, meaning that the coefficients from regressions (2), (3) and (4) only indicate how they changed compared to the coefficient in regression (1).

+ 𝛾₄∗(𝐶𝑎𝑠ℎ 𝑏𝑜𝑛𝑢𝑠)

(𝐵𝑎𝑠𝑒 𝑆𝑎𝑙𝑎𝑟𝑦)_𝑖𝑡∗ 𝐷𝐹 + 𝛾5∗

(𝑇𝑜𝑡𝑎𝑙 𝐸𝑞𝑢𝑖𝑡𝑦) _𝑖,𝑡∗ 𝐷𝐹 + 𝛾6∗(𝐷𝑜𝑙𝑙𝑎𝑟 𝑔𝑎𝑖𝑛 1%)𝑖𝑡∗ 𝐷𝐹 (2)

The second regressions row test the change in coefficient for non-financial firms after the Dodd-Frank Act. For example, a negative coefficient for 𝛾4 indicates that the relation between bonus

compensation over base salary and firm performance decreased after the Dodd-Frank Act for non-financial firms. However, this does not necessarily mean that the total relation after the Dodd-Frank Act is negative, since the total relation for non-financial firms is 𝛾1+ 𝛾4 for the short-term incentive

coefficient. This applies for all the three coefficients in the row.

+ 𝛾7∗

(𝐶𝑎𝑠ℎ 𝑏𝑜𝑛𝑢𝑠)

(𝐵𝑎𝑠𝑒 𝑆𝑎𝑙𝑎𝑟𝑦)𝑖𝑡∗ 𝐹𝐼 + 𝛾8∗

(𝑇𝑜𝑡𝑎𝑙 𝐸𝑞𝑢𝑖𝑡𝑦) 𝑖𝑡∗ 𝐹𝐼 + 𝛾9∗ (𝐷𝑜𝑙𝑙𝑎𝑟 𝑔𝑎𝑖𝑛 1%)𝑖𝑡∗ 𝐹𝐼 + (3)

The third regression row test the change in coefficients before the Dodd-Frank act for financial firms versus non-financial firms. The same applies for these coefficients as it does for regression in row (2). To test the difference in effect between financial and non-financial firms before the Dodd-Frank Act, one can analyze the coefficients 𝛾7 for the short-term incentive change or 𝛾8 & 𝛾9 for the long-term

incentive change. Again, these three coefficients only indicate the change compared to non-financial firms before the Dodd-Frank Act. The total relation for financial firms before the act is 𝛾1+ 𝛾7.

(18)

18 + 𝛾10∗ (𝐶𝑎𝑠ℎ 𝑏𝑜𝑛𝑢𝑠) (𝐵𝑎𝑠𝑒 𝑆𝑎𝑙𝑎𝑟𝑦)𝑖𝑡∗ 𝐹𝐼 ∗ 𝐷𝐹 + 𝛾11∗ (𝑂𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝 𝐶𝐸𝑂) (𝑇𝑜𝑡𝑎𝑙 𝐸𝑞𝑢𝑖𝑡𝑦) 𝑖𝑡∗ 𝐹𝐼 ∗ 𝐷𝐹 + 𝛾12∗ (𝐷𝑜𝑙𝑙𝑎𝑟 𝑔𝑎𝑖𝑛 1%)𝑖𝑡∗ 𝐹𝐼 ∗ 𝐷𝐹 (4)

Finally, regressions row four examines the effect for financial firms after the Dodd-Frank Act. Again, the coefficients in this row are an indicator of change in relationship between the incentive

coefficients and performance. As stated in the hypothesis section, we would expect significant positive coefficients for these three variables, since the purpose of the Dodd-Frank Act is to improve this alignment between CEOs and shareholders. It is sufficient enough to test only these coefficients, since we are interested in the change of relationship and are less interested if the total effect is positive or negative. Statistical significant coefficients in row (4) indicate that the Dodd-Frank Act accomplished one of its goals.

+ 𝛾13∗ 𝑆𝑡𝑜𝑐𝑘𝑅𝑒𝑡𝑢𝑟𝑛𝑖𝑡+ 𝛾14∗ 𝐵𝑜𝑜𝑘 − 𝑇𝑜 − 𝑀𝑎𝑟𝑘𝑒𝑡𝑖𝑡+ 𝛾15∗ 𝐿𝑜𝑔(𝑚𝑎𝑟𝑘𝑒𝑡𝑣𝑎𝑙𝑢𝑒)𝑖𝑡 ( 5)

This final regressions row contains the control variables used in this study. The coefficient itself are not interesting, meaning it does not matter for this study whether or not the coefficients are positive or negative. The control variables are included to prevent endogeneity problems in the form of omitted variables bias.

It is important to note that although the regressions rows are split above, the total model contains all regressions. The total regression model looks large, but this is due to the use of dummy interaction variables. The variables of interest will be explained briefly here in this section. A more in-depth clarification of the variable and the corresponding results will be discussed in section 4. ‘Cash

bonus/Salary’ controls for additional risk taking on the short term. The ‘Ownership(%)’ controls for

the long-term incentive the CEO has to maintain a high stock price. The higher the ownership of the CEO is, the more likely his long-term incentive will be aligned with the shareholders. The ‘Dollar gain

1%’ controls for a similar incentive problem as the ownership variable does. It indicates how much

the wealth of the CEO changes when the stock price goes up with one percent. While the change in stock price is in percentages, the change for CEO wealth is calculated in dollar value.

The regression will be done in a build-up style, which means every dependent variable will be regressed on its own and the next regressions will include more and more variables, until the point where the full model is being regressed and tested. Important to note is that the regression is a panel-regression. It will include time fixed effects and firm fixed effects, to control for additional non-observable variables. This will solve some of the endogeneity problems that we may encounter in datasets like these. The added control variables will also reduce the endogeneity issues, especially when using firm performance as an independent variable. Studies before show that factors such as the market value and the book-to-market value of a firm are highly correlated with firm

(19)

19 performance. Last year’s performance also influences the current year’s performance, as most of the time when a firm had a positive performance, the consecutive year will also most likely be positive.

When testing this regression, it is important to realize that the coefficients found in row two, three and four are showing the additional effect with respect to row one. This means that even when coefficients in row two are positive, it does not mean that the total effect of non-financial firms after the Dodd-Frank Act is positive. It only indicates that after the Act, the relation increased by a certain number, given the coefficients are significantly different from zero. The period of interest is between 2006 and 2014 for the compensation part. The year 2015 cannot be included because it will cause problems with calculating the stock returns for the next year. For the year 2006, only firms where the fiscal year ends on or after 15 December 2006 will be included. This is due to the fact that the SEC adopted new disclosure rules in 2006.

The alternative approach for this model as robustness check includes the use of return on assets and return on equity. The independent variables for the model will not change, but the dependent variable (stock returns) is replaced by either return on assets or return on equity. This alternative approach is used to test whether or not we find the same results as the original model. If the results are similar, we can conclude that the test results from the regression model are probably correct.

Following the methodology of Fahlenbrach & Stulz (2010) we analyzed the following model. First, when looking at the independent variables, the variables Equity Risk ($) and Equity Risk (%) are not included for this study. This is due to the fact that stock volatility is not party of my research, which was a point of discussion at the paper of Fahlenbrach & Stulz (2011). Secondly, the control variable for Tier 1 capital ratio is excluded from the methodology. This research takes into account the financial sector as a whole and not just the banks. Although the control variable is significant in their paper, banks only cover a portion of the total firms in the financial sector. The research paper from Fahlenbrach & Stulz (2011) is the foundation for the methodology that will be used. Before moving on towards the methodology, there will be a brief summary of how the model of Fahlenbrach & Stulz (2011) is structured and implemented. In the paper the academics look at the interest

alignment of bank CEOs before the crisis in 2006. For this, they analyze what fraction of the

compensation was paid in cash and what fraction was paid in equity. Next, they make a split for the cash part between basis salary and cash bonus. For the equity part, they distinguish between the fraction of stocks the CEO owns divided by the total stock and the change of the equity value of compensation if the stock would change by one percentage. The dependent variable used is the buy-and-hold returns of the firm from July 2007 to December 2008. The used regression model is:

(20)

20 𝐵𝑢𝑦 − 𝑎𝑛𝑑 − ℎ𝑜𝑙𝑑 𝑟𝑒𝑡𝑢𝑟𝑛𝑠2008,𝑖 = 𝛽1∗ (𝐶𝑎𝑠ℎ 𝑏𝑜𝑛𝑢𝑠) (𝐵𝑎𝑠𝑒 𝑆𝑎𝑙𝑎𝑟𝑦)𝑖+ 𝛽2∗ 𝑂𝑤𝑛𝑒𝑟𝑠ℎ𝑖𝑝(%)𝑖+ 𝛽3∗ (𝐷𝑜𝑙𝑙𝑎𝑟 𝑔𝑎𝑖𝑛 1%)𝑖+ 𝛽4∗ 𝐸𝑞𝑢𝑖𝑡𝑦 𝑟𝑖𝑠𝑘 ($)𝑖+ 𝛽5∗ 𝐸𝑞𝑢𝑖𝑡𝑦 𝑟𝑖𝑠𝑘 (%)𝑖+ 𝛽6∗ 𝑆𝑡𝑜𝑐𝑘𝑅𝑒𝑡𝑢𝑟𝑛 2006+ 𝛽7∗ 𝐵𝑜𝑜𝑘 − 𝑡𝑜 − 𝑚𝑎𝑟𝑘𝑒𝑡𝑖+ 𝛽8∗ 𝐿𝑜𝑔(𝑀𝑎𝑟𝑘𝑒𝑡𝑉𝑎𝑙𝑢𝑒)𝑖+ 𝛽9∗ 𝑇𝑖𝑒𝑟 1 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑅𝑎𝑡𝑖𝑜

Their findings, which have been mentioned before, find a significant negative coefficient for 𝛽3when

including all variables except 𝛽2 and 𝛽5. The control variables, 𝛽7 and 𝛽9 were also significant.

3.2 Data

The executive compensation data is retrieved from COMPUSTAT – EXECUCOMP. The database gives us a full detailed list of the base salary, cash bonuses, equity bonuses, stock options and more. For the financial industry, it is important to classify SIC codes 6000-6300 as such. COMPUSTAT also provides the data needed for the control variables. The Fundamentals Annually database gives an overview of the balance sheet items that we need, such as book value, market value and net income of the firms. The market value is calculated by taking the closing year price times to closing year outstanding stocks plus total assets on the balance sheet. For the dependent variable, CRSP offers monthly buy-and-hold returns, which will be used to calculate the yearly buy-and-hold stock returns of the companies. The other way to calculate the returns is by looking at the stock price at t=0 and calculate the relative change at t=1. The main disadvantage of this method is that it does not include monthly dividends that have been paid out.

For the EXECUCOMP data, I needed to make some alterations. First, we only have to keep the compensation data of the executives that have been marked by EXECUCOMP as “CEO”. This dropped the total number of observations from 108,755 to 16,260 observations. Secondly, to make sure there are no problems with the actually numbers, I removed the entries that did not report a base salary, bonus compensation, stock value in dollars and percentage stock owned. After this alteration, the total amount of observations went from 16,260 to 11,813. Additionally, the entries that reported negative numbers for the compensation have been removed as well. For the stock data from CRSP, I used a similar approach. If the data was missing, the entry has been removed. Finally, to make sure I only work with complete datasets, I removed the firms that had missing values in the Fundamentals Annually from COMPUSTAT. More specific, the firms that did not report total assets, net income, shares outstanding and closing share price had been removed. After this, the observation of the merged data set went from 11,812 to 7,731 observations.

(21)

21 Because the data set includes outliers, this study winsorized some of the dependent variables at a level of five percent. This same technique is used by Fahlenbrach & Stulz (2011) where they winsorize the one percentage increase in stock value at a level of two percent and the other variables at a level of five percent. This study mainly uses this technique for the stock returns, return on assets and return on equity.

3.3 Descriptive statistics

This section covers the descriptive statistics retrieved from the modified database. The data regarding the executive compensation will be split up between base salary in dollars, bonus

compensation in dollars, stock owned in dollars and stock owned in percentages of total outstanding stocks. Within each table the data is separated between non financials/financials and separated between a period before the Dodd-Frank Act and after the Dodd-Frank Act.

Table 1 contains the base salary of the North America CEOs in thousands of dollars. Overall, the base salary is not different between non-financials and financials CEOs. The total average is 731,036$ for the total time period. However, there are some interesting numbers to point out in this table. For both the non-financial and financial firms, the base salary increased over time from

665,257$ to 790,249$. Additionally, the increase is relatively, and in absolute numbers, higher for the non-financials than it is for the financials. The average base salary found by Fahlenbrach & Stulz (2011) is 761,000$, slightly different than the numbers in table 1. The first difference is due to the fact that Fahlenbrach & Stulz only look at banks, where financials in table 1 include all firms in the financial sector. Secondly, the difference is caused by the different time periods used, where Fahlenbrach & Stulz only look at the compensation of 2006 and Table 1 looks at the time from 2007 until 2014.

Table 2 provides the average bonus compensation found in thousands of dollars. Where the base salary in Table 1 was rather equal between non-financials and financials, we find a large difference between the bonus compensation earned by CEOs between non-financials and financials. There are two major differences that stand out in this table. First, the on average relatively decrease (-13.27%) of bonus compensation for financial firms after the Dodd-Frank Act and second, the difference between bonus compensation for the non-financials (167,102$) and the financials (341,280$). This may suggests that incentive based compensation is a much bigger deal in the financial sector than it is in the non-financial sector.

Table 3 and Table 4 present the descriptive statistics for the value and percentage stock owned by the CEOs. Table 3 provides the average dollar value for stocks owned by the CEOs. Just as we saw with the means in Table 1, the value overall has increased over the two time periods, from

(22)

22 $1339.87 thousand in the period before the Dodd-Frank Act to $2404.46 thousand in the period after the Dodd-Frank Act. This increase can be justified by the increasing stock prices over the same period. For example, the Dow Jones reached the 6,547.05 on March 9 2009, the lowest point during the crisis. From that point on, the index increased and closed on December 30 2014 on 17,983.07. The index almost doubled over the two periods, explaining the large differences between stock values for the CEOs.

Finally, Table 4 provides information about the total percentage stock owned by the CEO with respect to all the shares outstanding of the firm. It is important to look at both total value and percentage owned, since the two are not one-on-one related. The stock value owned by the CEO might go up, despite the fact the total percentage shares owned goes down. This scenario is exactly what we find in the data. There is a decrease overall in percentage stocks owned by the CEOs for all industries, although Table 3 shows us an increase in value of the same stocks. The percentage stock owned in the non-financials industry is significantly higher than the percentage stock owned in the financial sector, respectively 3.1128% and 1.7651%. This contradicts the statement previously made that CEOs in the financial sector receive more incentive/equity based compensation.

Remaining descriptive statistics can be found in the appendix (A). Variables included in the appendix are firm’s total assets (at), net income (ni), bonus compensation / base salary (bs_ratio) and the market-to-book ratio (mtb_ratio). Without going in to too much details, some descriptive statistics are important to note. The average net income between non-financials and financials is around the same. However, the average total assets of the non-financials (8,622,000$) is significantly lower than the average total assets of the financial firms (48,791,000$). This large difference

between the two is caused by the banks that are included in the data sample. Banks such as JP Morgan (2,415,689,000$) and Bank of America (2,264,909,000$) increase the mean of total assets significantly. This is not very surprising, since banks have much higher total assets due to the way the balance sheets are constructed. Another thing to point out is the difference in market-to-book ratio between non-financials (2.4071) and financials (1.6626). Again, this is caused by the high book value of the financial firms, which makes is unlikely that the ratio would be very high. At last, the

bonus/salary ratio between non-financials and financial firms is large. The total average bonus/salary ratio for non-financial firms (5.4435) is significantly higher than the ratio for the financials (0.5497).

Appendix (B) provides the correlation statistics on the important variables used. It is

interesting to note that correlation between the returns and bonus compensation are fairly low. This is due to the fact that the returns included in the correlation matrix are next year’s returns. Finally appendix (C) includes a graphical view of the change of CEO compensation over the period of time.

(23)

23 Table 1 Summary of base salary ( # of firms) in thousands of dollars

Time Period Non Financials Financials All Firms

Before Dodd-Frank 2007-2010 664.156$ (3247) 674.979$ (912) 665.527$ (4159) After Dodd-Frank 2011-2014 793.698$ (3565) 777.519$ (966) 790.249$ (4531) Total Time Period 2007-2014 731.951$

(6812)

727.718$ (1878)

731.036$ (8690)

Table 2 Summary of bonus in thousands of dollars

Before Dodd-Frank 2007-2010 162.705$ 366.299$ 207.350$

After Dodd-Frank 2011-2014 171.107$ 317.661$ 202.352$

Total Time Period 2007-2014 167.102$ 341.280$ 204.744$

Table 3 Summary of stock owned in percentages

Before Dodd-Frank 2007-2010 3.9204% 2.2461% 3.5533%

After Dodd-Frank 2011-2014 2.3773% 1.3110% 2.1499%

Total Time Period 2007-2014 3.1128% 1.7651% 2.8216%

Table 4 Summary of stock owned in thousands of dollars

Before Dodd-Frank 2007-2010 1306.62$ 1458.27$ 1339.87$

After Dodd-Frank 2011-2014 2394.78$ 2440.18$ 2404.46$

(24)

24

4. Analysis

This section provides the empirical results found from the regression model. Section 4.1 uses the original model, where the dependent variable is the yearly stock returned. The regressions is split into multiple sections and test the difference between the results found in this study and the results found by Fahlenbrach & Stulz (2011). Section 4.2 presents the results of the robustness check, where two alternative dependent variables are used to test the model.

4.1. Empirical Results

Table 5 shows the results found from running the eight different regressions. Regressions (1) includes Bonus/Salary independent variable as main explanatory variable. Regressions (2) includes percentage of stock owned by the CEO as part of the total stock outstanding. Regressions (3) includes the one percent change in stock price on the stock value of the CEO as independent variable. Regressions (4) and (5) use the approach done by Fahlenbrach & Stulz (2011), where they chose to use either percentage stock owned as an additional explanatory variable or the change in stock value as an additional explanatory variable. Regressions (6) to (8) include all independent variables. The main difference between the three regressions is the use of fixed and year effects, and the addition of the three control variables. As stated before, the regressions are created in a build-up style, meaning the first regressions contain only the independent variables and the last regression contains all variables including the control variables.

As stated before, the first regressions includes only the bonus compensation over base salary ratio as an indicator for the high short-term incentives of the CEO. Similar to findings from

Fahlenbrach & Stulz (2011), the coefficient is negative and significantly different from zero. This indicates that, because it reflects short-term incentive of the CEO, CEOs that receive high bonuses and low base salary have lower stock returns. Additionally, the second coefficient is also negative and statistical significant. This means that the coefficient for non-financial firms after the Dodd-Frank Act changed significantly, from -0.00513 to -0.01004. The total coefficient decreased with -0.00491 after the Dodd-Frank Act was implemented. Although the coefficients for the financial industry are very high, +1.71715 before the Dodd-Frank Act and +0.86389 after the Dodd-Frank Act, we can’t accept the hypothesis that they are significantly different from zero due to the low T-values.

(25)

25

Table 5 Dependent variable: Yearly Stock Return

(1) (2) (3) (4) (5) (Bonus/Salary) -0.00513*** -0.00557*** -0.00523*** (-4.02) (-4.00) (-4.06) (Bonus/Salary) * DF -0.00491*** -0.00367*** -0.00485*** (-19.70) (-8.00) (-18.85) (Bonus/Salary) * FIN 1.71715 1.81052 2.50321* (1.21) (1.29) (1.71) (Bonus/Salary) * FIN * DF 0.86389 0.61789 -0.03379 (1.27) (0.92) (-0.05) % Stock Owned 0.18985 0.18886 (0.77) (0.76) % Stock Owned * DF -0.34148*** -0.33348*** (-3.15) (-3.06)

% Stock Owned * FIN -0.96944* -0.88627

(-1.74) (-1.63)

% Stock Owned * FIN * DF 0.99248** 0.97291**

(2.48) (2.41) 1% Stock Delta 0.00028 0.00029 (1.25) (1.28) 1% Stock Delta * DF -0.00026 -0.00027 (-1.18) (-1.21) 1% Stock Delta * DF -0.00194*** -0.00207*** (-3.85) (-4.31)

1% Stock Delta * FIN * DF 0.00205*** 0.00218***

(4.63) (4.84)

Log(Total Assets) -6.78843** -6.65529** -6.63459** -6.64863** -6.51762**

(-2.36) (-2.32) (-2.27) (-2.35) (-2.26)

Market-to-book 24.82185*** 24.8763*** 24.8562*** 24.8619*** 24.8349***

(17.50) (17.51) (17.54) (17.49) (17.52)

Stock Return prior year -0.31609*** -0.31780*** -0.31762*** -0.31775*** -0.31789***

(-28.98) (-29.21) (-29.14) (-29.15) (-29.07)

Fixed Firm Effects YES YES YES YES YES

Fixed Year Effects YES YES YES YES YES

N 7731 7731 7731 7713 7731

R-Squared (Within) 0.4662 0.4668 0.4673 0.4672 0.4679

R-Squared (Overall) 0.1833 0.1824 0.1825 0.1843 0.1855

Includes the buy-and-hold returns as dependent variable and proxies for executive compensation as independent variable. The table shows the results from a panel regression of the buy-and-hold stock returns on the CEO’s 1. Bonus compensation over base salary 2. Percentage stock owned with respect to total shares outstanding 3. The increase in dollar value at a one percent increase in stock price. “Bonus/salary” is the bonus compensation as the cash bonus earned by the CEO and provided by EXECUCOMP, divided by the base salary of the CEO in the same year. “% Stock owned” is the total amount, restricted and unrestricted stocks, owned by the CEO divided by the total amount of shares, restricted and unrestricted, outstanding times 100%. “1% Stock delta” is the increase in stock value of the CEO when the stock price increases by one percent. Numbers for book value and market value are all based on prior year firm values. In parentheses the T-values by given by STATA are reported. The statistical significance is indicated by *, **, *** which are alpha levels for 10%, 5% and 1%, respectively.

(26)

26 An important thing to note when comparing column (1) to (5) with each other is that the R-squared value, both within and overall, are roughly the same. This is due to the fact that the explanatory power of the fixed effects, year effects and control variables is high. The T-values for both market-to-book ratio, prior year stock return and year effects are large. These findings are an important difference with the findings of Fahlenbrach & Stulz (2011), where they find no statistical significance for the prior stock returns and for the log of market value.

Column (2) examines the relation between stock return and percentage stock owned by the CEO in his/her firm. Fahlenbrach & Stulz (2011) do not find a significantly different from zero coefficient. The results shown in Table 5 suggest the same, however, this only implies for non-financial firms before the Dodd-Frank Act. After the Dodd-Frank Act we find a significant negative coefficient of -0.34148. Because the first of the four coefficients is not significant, we cannot simply add the two coefficients together to see the total effect of stock percentages owned for the non-financials. We can only conclude that, whatever the effect was before the Dodd-Frank Act, it decreased in the period after the act. For financial firms there is a negative coefficient before the Dodd-Frank Act (-0.96944), but a positive effect after the Dodd-Frank Act (+0.99248). Again, the same as for the non-financials, we cannot assume that the total effect for a financial firm before the Dodd-Frank Act is equal to the coefficient (-0.96944), since that coefficient is an addition on the first coefficient, which is insignificant. It is interesting to note that the coefficient is so vastly different for the financials before and after the act. The economic meaning of this is that CEOs before the act holding a large percentage of the total stock of their company performed worse than the non-financials. However, after the Dodd-Frank Act, the CEOs of the financials were performing better when they had more percentage stock of their firm. The coefficient for the financials improved where the coefficient for the non-financials decreased.

Column (3) includes the one percent increase in total CEO stock value as independent variable. All four coefficients are relatively small and the coefficients corresponding with the non-financials are insignificant as well. Nevertheless the coefficients for the non-financials are both significantly strong (P-value under 0.01) and show a similar results as column (2) meaning the coefficient went from negative before the Dodd-Frank Act to positive after the act. This could indicate that in the period after the act, the incentives of the CEOs in the financial sector were better aligned. Finding the causal impact of the Dodd-Frank Act is difficult to prove, but it seems the

variables for percentage stock owned and for the one percentage increase in stock value the alignment between interest of the CEO and interest of the shareholders became better.

(27)

27 The results from column (4) and (5) are not different than the previous regressions. The only

differences come from the level of significance for the bonus over salary coefficient for financial firms before the Dodd-Frank Act in column (5), which goes from insignificant to significant at an alpha of 10%.

Table 6 Dependent variable: Yearly Stock Return

(6) (7) (8) (Bonus/Salary) 0.00254 0.00679*** -0.00556*** (0.60) (5.28) (-3.97) (Bonus/Salary) * DF -0.00506 -0.00349 -0.00373*** (-0.83) (-7.57) (-8.17) (Bonus/Salary) * FIN 0.68189 2.98728** 2.50157* (0.93) (2.36) (1.73) (Bonus/Salary) * FIN * DF 0.05848 0.13367 -0.12799 (0.48) (0.22) (-0.19) % Stock Owned -0.33581*** -0.16387 0.1847 (-3.29) (-0.69) (0.75) % Stock Owned * DF 0.14989 -0.2415** -0.31082*** (1.03) (-2.21) (-2.86)

% Stock Owned * FIN -0.93822** -0.23727 -0.89831

(-2.33) (-0.42) (-1.60)

% Stock Owned * FIN * DF 1.1214* 0.43716 0.65873

(1.83) (1.38) (1.61) 1% Stock Delta 0.00079*** 0.00007 0.00031 (3.07) (0.27) (1.38) 1% Stock Delta * DF -0.00071*** -0.00003 -0.00029 (-2.67) (-0.11) (-1.32) 1% Stock Delta * DF -0.00188 -0.001963*** -0.00203*** (-3.36) (-4.21) (-4.26)

1% Stock Delta * FIN * DF 0.00181*** 0.00151*** 0.00203***

(2.78) (3.36) (4.47)

Log(Total Assets) -6.43019**

(-2.27)

Market-to-book 24.8679***

(17.50)

Stock Return prior year -0.31918***

(-29.24)

Fixed Firm Effects NO YES YES

Fixed Year Effects NO YES YES

N 7731 7731 7731

R-Squared (Within) 0.0050 0.3189 0.4687

R-Squared (Overall) 0.0057 0.2800 0.1856

Includes the buy-and-hold returns as dependent variable and proxies for executive compensation as independent variable. The table shows the results from a panel regression of the buy-and-hold stock returns on the CEO’s 1. Bonus compensation over base salary 2. Percentage stock owned with respect to total shares outstanding 3. Numbers for book value and market value are all based on prior year firm values. In parentheses the T-values give by STATA are reported. The statistical significance is indicated by *, **, *** which are alpha levels for 10%, 5% and 1%, respectively.

(28)

28 Column (6) to (8) in table 6 show the importance of including control variables to the model, including fixed effects to the model and including time effects. The best way to illustrate this is by looking at the R-squared, both within as overall, which are increasing from regressions (6) to 7 and increase from (7) to (8). Some economic meaning can be derived from column (8). For the non-financials, the coefficient for bonus compensation of base salary is negative, meaning that before the Frank Act CEOs with high bonuses and low base salary performed badly and after the Dodd-Frank Act these group of CEOs performed even worse. The non-financial CEOs that owned more percentage of their own firm’s stock performed badly in the period after the Dodd-Frank Act. At last, column (8) shows us the same relation as we saw in column (3) and column (5). A negative coefficient before the Dodd-Frank Act moving towards a positive coefficient after the Dodd-Frank act.

As noted before, the R-squared results are similar between all five columns. For all of the five regressions, the Log (Total Assets) is significant at an alpha of 5% and both the market-to-book ratio and prior year stock return are significant at an alpha level of 1%. Because all the results are so close to each other, we cannot conclude which model is best to explain the research question. They all have in common that for the short-term incentive (bonus/salary) the coefficient decreased for non-financials. Secondly, it seems that for the financials the long-term incentive coefficient increased after the Dodd-Frank Act, for both percentage stock owned and one percent increase in stock value. The negative coefficient for prior year stock return is surprising, since Fahlenbrach & Stulz (2011) found an insignificant result in their paper. Nonetheless, in the paper of Beltratti & Stulz (2009), they do find a statistical significant negative relation between the prior year returns and the current year returns for international banks. Another major difference in statistical evidence in the control variable is the difference between coefficients for market-to-book. Where this thesis finds a positive significant coefficient, Fahlenbrach & Stulz (2011) find a negative relation between market-to-book and firm’s stock returns.

4.2 Robustness check

Where the regressions in section 4.1 focus on the yearly stock returns as dependent variable, we now use account performance measures as dependent variable into the regression. Both return on assets and return on equity will be used as the alternative approach to the stock returns. The buildup of this section will be similar to the one in section 4.1. Column (1) to (4) regarding the return on assets will be discussed as first then column (5) regarding the return on equity will be discussed and at last the overall statistical and economical value of the new approach will be discussed and compared with the normal method.

The effect of the Dodd-Frank Act on the misalignment of incentives between CEOs and shareholders