The impact of corporate governance on bondholder value

(1)

University of Amsterdam

The Impact of Corporate Governance on

Bondholder Value

Yuanchun Li (10174028)

Master Thesis Business Economics Finance Track

Faculty Economics and Business Supervisor: Tolga Caskurlu

(2)

Statement of Originality

This document is written by Yuanchun Li, 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)

Abstract

This thesis investigates the effect of corporate governance on bondholder value in S&P 1500 firms by applying a combination of the regression discontinuity design and the event study. The regression discontinuity design is able to capture the causal relationship between passing shareholder proposals and bond abnormal returns and overcome the endogenity problem that is pervasive in previous studies. The empirical results in this thesis indicate that passing shareholder proposals has negative effects on bond abnormal return during the period 2002 to 2013. In addition, the effect is positive in firms with event risk covenants as well as in non-investment graded.

(4)

raised due to the separation of ownership and control (Coase, 1937). Investors invest money in companies in order to get returns and managers get financing from investors for companies’ future development, however, it is difficult for investors to make sure that their funds are not wasted by managers and that is called the agency problems. Corporate governance is invented to solve the agency problem and protect investors’ rights and improve firm performance. Managers in firms with strong corporate governance are assumed to have aligned interests with shareholders (Jensen & Mecking, 1976).

A lot of research has been done on the relationship between corporate governance and shareholder value and find that higher corporate governance quality increases stock returns (Gompers, Ishii & Metrick, 2001; Cremers & Nair, 2005). While most previous studies focus on the impact of corporate governance on stock performance and shareholder value, the impact of corporate governance on bondholders has not been investigated intensively yet. As discussed by Jensen and Meckling (1976), there are two different levels of conflicts of interest. The first one is the conflicts between managers and all external stakeholders, and the second one is the conflicts between shareholders and. A large amount of research shows that shareholders can benefit from the improvement of corporate

governance; however, whether the bondholder can benefit from strong corporate shareholder governance is not clear. Therefore, this thesis aims to analyze the impact of corporate governance on bond abnormal return to analyze the corporate governance effect on bondholders.

Previous studies mainly use OLS regression as the study method, while this method is subject to two problems as discussed by Cunat, Gine and Guadalupe (2012). First, corporate governance is endogenous and correlated with other firm characteristics. Second, investors’ knowledge of the firm should already be shown in the market price, which means there should not be any abnormal return. These two problems make it very difficult to estimate the pure causal

(6)

relationship between corporate governance and abnormal returns. To solve these two problems, this thesis adopts the regression discontinuity design proposed by Cellini, Ferreria and Rothstein (2010) and Cunat, Gine and Guadalupe (2012). Since the probability of implementing a proposal is discrete around the threshold between firm passed a proposal by a small margin and firms rejected it by a small margin, the regression discontinuity design is able to capture the causal effect of passing a shareholder proposal on bond abnormal return, as long as the other firm characteristics are continuous around the threshold. Furthermore, previous studies mainly use bond yield spread in their analysis while I choose to use bond abnormal return since the abnormal return can capture the effect much better. Thanks to the introduction of TRACE, it makes the calculation of bond abnormal return feasible since it reports daily bond prices. In addition, I also add some other dummy variables, such covenant and non-investment graded into my analysis in addition to the model used by Cunat, Gine and Guadalupe (2012). This analysis will not only enrich the literature in this area but also provide

directions to the future development of corporate governance in the bondholders’ respective.

Due to the data availability, the study will cover the period 2003-2013. The shareholder proposal data can be obtained from Institutional Shareholder Service (formerly RiskMetrics), bond price and volume data is from TRACE and the bond covenants and rating data is from FISD. Treasury information is from the US department of treasury website. The shareholder proposal will be put into six different categories (Cunat, Gine and Guadalupe, 2012), namely, audit, board, compensation, G-Index (antitakeover index), voting and other. According to previous literature, antitakeover plays important role corporate governance; therefore special attention will be paid on the G-index type proposals. Other firm characteristics variables are collected from Compustat.

I find that passing a proposal has negative impact on bond abnormal returns. I also find that the G-index proposal has negative impact on abnormal return as well, even though previous studies find that anti-takeover provision can benefit bondholders. Since the use of event risk covenants align the interest of bondholders and shareholders, the impact of passing a proposal on bond

(7)

abnormal return is positive in firms with event risk covenants. The impact is also positive in non-investment graded firms. My analysis also indicates that the more proposals passed the larger the negative impact on abnormal return.

The rest of this thesis is organized as follows. The background information and a literature review are given in next section. Then in section 3 the data and descriptive statistics are reported. The methods and models used in this thesis’s analysis are explained in section 4. Section 5 presents the results and comes with the conclusion section in the end.

(8)

2. Background and literature review 2.1 Background

As the company is growing bigger, they are looking for external finance source and that is when the investors appear. Since investors and managers are different entity, there may have different interests in this company and those interests may be conflicted with each other and this is called the agency problem. The agency problem is raised due to the separation of ownership and control (Coase, 1937). Investors invest money in order to get returns and managers get financing from investors for companies’ future development, however, it is difficult for investors to make sure that their money are not wasted by managers. There are two types of agency problems (Jensen and Meckling, 1976). The first one is the conflicts between shareholder and managers. Shareholders expect to have higher return while managers may only focus on their private benefits. To solve this problem, corporate governance is invented to protect shareholders and improve firm performance. While Jensen and Meckling (1976) indicate that mangers in firms with strong corporate governance are assumed to have aligned interests with shareholders, there are not many studies focus on the interests of bondholders. The second agency problem is the conflicts between shareholder and bondholders. Strong shareholders could take advantage of bondholders by, for example, taking very risky projects.

Shareholder proposal is a kind of shareholder activism and it is considered as a way to improve shareholder governance. Investors submit shareholder proposal to give recommendation on the company’s future action. However, the outcome of the vote is not binding, which means even the proposal passes, the company may still not implement it and the proposal may still be implemented even though it is not approved. According to Ertmur, Ferri, and Stubben (2010), only 31.1% of the shareholder proposals that pass are implemented, while only 3.2% of the shareholder proposals that fail are implemented.

2.2 Literature review

There are mainly two types of empirical studies analyze the relationship between investors value and corporate governance. The first type of studies focuses on the effect of corporate governance and shareholder value; while the second type of

(9)

studies puts much attention on the influence of corporate governance on bondholder value. Most of the previous studies on both the first and the second type studies employ OLS regression with different control variables. However, they fail to find the causal relationship between investor value and corporate governance due to the endogenity problem. Only recently, the regression discontinuity design is introduced in the study in this area and that pushes researches in this area in another level.

2.2.1 Corporate governance and shareholder value

The first type of studies mainly focuses on the relationship between corporate governance and shareholder value. Since corporate governance is designed to solve the conflicts between investors and managers, the introduction and improvement of corporate governance are supposed to increase shareholder value. However, corporate governance also increases the level of freedom of managers in the same time, which allows the managers to pursuing private benefits and hurt shareholder value. Moreover, more and more firms start to implement anti-takeover provisions, which are known as bad for shareholders. Even though there are many different ways to measure shareholder value, the main focus is on analyzing the impact of corporate governance on firm performance (mainly Tobin’s Q) and stock abnormal returns. Most of the previous studies test the relationship between corporate governance and shareholder value using OLS regression and setting governance index as the independent variable(s) and Tobin’s Q/stock abnormal return as the dependent variable. In 1999, Gillan and Starks show that the market reaction to corporate governance depends on different types of governance proposal. Institutions sponsored proposal has positive impact on stock prices while individual sponsored proposal has negative but not significant impact on stock prices. By using corporate governance rating data, Drobetz, Schillhofer and Zimmermann (2004) find positive relation between corporate governance and firm valuation but expected stock returns has negative relation with corporate governance. Masulis, Wang and Xie (2007) find that more anti-takeover provisions lead to lower abnormal return since anti-takeover provisions diminish the corporate control on managers and destroy shareholder value. In 2003, Gompers, Ishii and

(10)

Metrick introduced the well-known corporate index—G-index. They count the number of corporate governance provisions that are considered hurt shareholder’s benefits to measure the level of corporate governance in a firm. They find that better governance improves shareholder value and brings better firm operating performance. Similar to Gompers, Ishii and Metrick (2001), Black, Jang and Kim (2006) find additional support for the findings that better corporate governance increases firm value. Cremers and Nair (2005) find that since in 1990 investors did not pay much attention on the gains of good governance, therefore the gain from a portfolio that buys firms with the highest level of takeover vulnerability and shorts firms with the lowest level of takeover vulnerability may create an abnormal return of 10% to 15%. However, they fail to provide pure causality relationship; therefore, some hypotheses cannot be explained in their study. Recently, Cunat, Gine and Guadalupe (2012) find a way to analyze the causal relationship between corporate governance shareholder proposal and stock abnormal return by using regression discontinuity design and event study. They claim that passing a shareholder proposal leads to magnificent positive abnormal returns.

As can be seen, even though a vast number of researches have been done on the corporate governance-shareholder value relationship; there were no general results of these empirical studies. However the majority results show that strong corporate governance has positive impact on stock and firm performance. All those findings are valuable contribution to the corporate governance literature, but they pay no attention on how the change of corporate governance structure might impact the bondholder value.

2.2.2 Corporate governance and bondholder value

Given the vast number of empirical studies that attempts to analyze the impact of corporate governance mechanisms on stock abnormal returns and firm value, more attention is deserved to be paid on the relationship between bondholder value and corporate governance. As a consequence some preliminary research has been conducted in this area. Since stronger corporate governance can reduce agency problem and provide better monitoring, therefore, information asymmetry and default risk are reduced. As a result, the improvement of

(11)

corporate governance lowers the cost of debt (Bhojraj & Sengupta, 2003). To be more specific, they find some empirical evidence showing that firms with higher institutional ownership and stronger outside control have a lower cost of debt financing. Cremers and Nair (2005) investigate the impact of both external and internal corporate governance and find that only when the company has high quality external corporate governance, high internal corporate governance quality leads to significantly positive stock performance. Anderson, Mansi, and Reeb (2003) use ownership structure to analyze the relationship between corporate governance and cost of debt. Since firms with family ownership have high reputation among the shareholder and bondholders, and also family owners tend to take their responsibility of monitoring actively so that the agency problem is reduced, thus, firms with family ownership usually have lower cost of debt. However, they only focus on the ownership structure and fail to analyze corporate governance in other ways rather than ownership structure.

Anti-takeover provisions may depress stock prices and sometimes managers use anti-takeover provision to entrench themselves in the company. However, Klock, Mansi and Maxwell (2005) find that anti-takeover provisions are beneficial for bondholders, showing that that corporate governance impacts stockholders and bondholders in opposing ways. Since bond covenants restrict managers from engaging in obtaining private benefits therefore the protection of bondholders’ improved (Defond & Jiambalvo, 1994) and the likelihood of default is decreased (Watts & Zimmerman, 1986). Cremers, Nair and Wei (2007) show that the use of event risk covenants reduces the credit risk that is brought by strong shareholder governance and without covenant, the interests of shareholder and bondholder diverge.

2.3 Hypotheses

The hypotheses of this thesis are based on the literature review above as well as the economic theories about the agency problem.

H1: passing a proposal in general leads to decline of bond abnormal return.

This hypothesis tests the second agency problem (see section 2.1). A shareholder proposal mainly considers the interests of shareholders instead of the interest of

(12)

bondholders. With the existence of the second agency problem, shareholders may take advantage of bondholders. Both studies from Klock, Mansi and Maxwell (2005) and Cremers, Nair and Wei (2007) find that the interest of shareholders and bondholders diverge.

H2: passing a proposal leads to increase of bond abnormal return in firms with event risk covenants.

The presence of event risk covenants reduces the credit risk associated with strong shareholder governance and aligns the interests of shareholder and bondholder together (Cremers, Nair and Wei, 2007). Therefore, with covenants, bond abnormal should be increased.

H3: passing a proposal leads to increase of bond abnormal return in non-investment graded firms.

Non-investment graded firms is suffering more risks compared with investment graded firms. In theory, passing a proposal in general improve the quality of corporate governance and decrease the default risk in this firm in general. Therefore, bondholders should also benefit from the improvement of corporate governance.

H4: passing G-index type proposals may improve bond abnormal return compared with other types of proposals.

G-index measures the number of anti-takeover provisions in place in a firm. As mentioned above, anti-takeover provisions may depress stock prices and sometimes managers use anti-takeover provision to entrench themselves in the company, anti-takeover provision may hurt shareholder value and increase bondholder value (Klock, Mansi and Maxwell, 2005).

H5: the more proposals passed in one meeting, the larger impact on bond abnormal return.

Following the way of creating G-index, I also create an index to capture the impact of the number of proposal passed. The larger the number of proposal passed, the higher the quality of corporate governance will be improved, therefore, the larger impact from corporate governance on bond abnormal return.

(13)

3. Data and descriptive statistics

This thesis combined several databases in order to get the complete dataset that is needed in the analysis. A description of all the database and variables used in this thesis will be given in this section. In addition, a brief description of how to prepare the dataset followed with the descriptive statistics will be presented as well.

3.1 Data

The Trade Reporting and Compliance Engine (TRACE) reports daily data in the Financial Industry Regulatory Authority (FINRA)’s over the counter bond market. The daily data includes bond price, bond yield and trade volume in each individual bond transactions. Since TRACE was first introduced in July of 2002, the sample in this thesis can only covers the period starts from July of 2002. Mergen Fixed Income Securities Database (FISD) provides additional data in the bond market. The data of covenant and investment grade is collected in FISD and I then merge them with TRACE manually. In order to calculate the abnormal returns of the bond, the treasury data is needed as well and I collect all the treasury data from the US Department of the Treasury website by hand. Combined TRACE, FISD and treasury data, I get the complete bond dataset for my thesis.

The shareholder proposal data is collected from Institutional Shareholder Service (ISS, formerly RiskMetrics). ISS provides both governance and social responsibility proposals in all S&P 1500 companies. ; I only focus on governance proposals in this thesis. The ISS variables that I used in this thesis include: company name, ticker, meeting date, vote percentage and proposal type. ISS also provides data about the G-index. G-index is introduced by Gompers, Ishii and Metrick in 2001 and it counts the number of different antitakeover proposals that is in place in the firm. ISS classifies the shareholder proposals into 72 different types, while this classification might introduce heterogeneous problems into the dataset as mentioned by Cunat, Gine and Guadalupe (2012). So I follow the categories that are created by Cunat, Gine and Guadalupe (2012) and group the proposals into six broader categories: audit, board, compensation, G-index, voting and other. Since G-index type proposals has more observations which meet the

(14)

Table 1

Shareholder Governance Proposals Summary Statistics

Data in this table are collected by ISS (Ristmetrics) on all shareholder governance proposals from 2002 to 2013 for all S&P 1500 companies. The threshold for approval is 50% for all besides 13 observations. These 13 observations are eliminated from the sample. Panel A indicates number of proposals, number of approved proposals, approved percentage and average voting percentage each year. Panel B classifies governance proposals by type.

Panel A: Shareholder Proposal Summary Statistics (by year) Year No. of Shareholder Proposal No. of Approved Proposals Percentage Approved Proposals Average Vote

Percentage Std. Dev. Vote Percentage 2002 23 6 26.09% 33.48% 21.0797 2003 233 74 31.76% 33.76% 22.8333 2004 220 53 24.09% 29.43% 23.1921 2005 264 76 28.79% 35.96% 23.6470 2006 316 98 31.01% 39.65% 21.9513 2007 324 71 21.91% 35.28% 20.6827 2008 244 63 25.82% 39.07% 21.6239 2009 321 79 24.61% 42.69% 20.0483 2010 291 79 27.15% 40.12% 20.4674 2011 201 70 34.83% 42.61% 21.4042 2012 219 70 31.96% 43.52% 23.3457 2013 209 51 24.40% 38.57% 23.2889

Panel B: Shareholder Proposal Summary Statistis (by broad classification) Proposal Type _ProposalsNo. of Approved No. of

Proposals

Percentage

Approved Average Vote Percentage Audit 17 0 0.00% 18.90% Board 673 94 13.97% 33.82% Compensation 939 92 9.80% 27.50% G-index 901 485 53.83% 52.28% Voting 302 128 42.38% 48.33% Other 133 7 5.26% 16.23%

analysis requirement in this thesis (see descriptive statistics), I further pool proposals in other five categories together for future analysis. Since the ISS database is only available until 2013, it determines the end of my sample period.

Besides the bond and shareholder proposal data, I collect additional information for all the firms from Compustat database. Compustat provides financial information for all S&P companies.

Most of the databases mentioned above are available in WRDS. The data is collected first and then merged manually or by using Stata. Since some databases

(15)

Table 2

Descriptive Statistics

The bond abnormal returns and bond yield spread are computed by using the date from TRACE and the US Department of the Treasury. G-index is introduced by Metrick et al (2001) and it is the number of antitakeover provisions in place at the firm. The G-index data is provided every two years by ISS. Covenant is the number of certain covenants in place at the firm. Four different types of covenants are considered so the max number of covenant is four. All other accounting variables are obtained from Compustat. Tobin's Q is defined as the market value of assets divided by the book value of assets, and balance sheet Deferred Taxes and Investment Tax Credit. Book to market is the ratio of book value of common equity to market value of common equity. Note the number of observations may change due to missing values for some of the variables.

N Mean Std. Dev. 10th Per. 90th Per. Bond abnormal return on meeting day 1119 -0.0232 8.23 -6.2013 9.9993 Bond yield spread 1026 0.9199 3.56 -1.9242 3.1470 G-index 1102 9.1343 2.32 6 12 Covenants 1119 1.1242 1.17 0 3 Total assets ($millions) 1114 134,758.50 336056.70 8,673 237,512 Market value ($millions) 1025 48,627.34 65040.16 4,987 138,757 Capital expenses ($millions) 993 2,611.10 4488.36 143 6,322 Tobin's Q 849 1.5730 0.73 0.9788 2.3859 Book to market 1025 0.3732 2.63 0.1762 0.8754 Return on equity 1001 0.4514 5.42 0.0179 0.3517

do not have unique identifiers for merging, I first merge them by using ticker and then check whether the merge is accurate manually. All the analyses are conducted by using the software package Stata.

3.2 Descriptive statistics

Table 1 displays the summary statistics about the shareholder proposals. Panel A of Table 1 shows the frequency of shareholder proposals and approved proposals, percentage of approved proposals, and mean and standard deviation of the proposals. Started from 2002, the average in favor of vote percentage is increasing and the percentage of approved proposals is rising as well. That could be a signal that corporate governance plays more important role nowadays. Panel B indicates that G-index proposal has the highest “passing rate” which is 53.83% and voting proposal takes the second place with 42.38% of proposals approved in the past eleven years. Consider the proposal quantity and percentage approved and average voting rate, G-index proposals are more suitable for my analysis. In order to get enough observations, I pool other five types of proposals together as a counterpart group of the G-index proposal.

(16)

Table 2 displays the characteristics of the firms in the sample. The bond abnormal return is calculated by using TRACE (see more in section 4). The bond yield spread is the difference between the bond yield and the matched treasure spread. G-index is the governance index introduced by Gompers, Ishii and Metrick (2001) and the maximum number is 16. I use four different types of covenants, namely, leverage covenant, networth covenant, poison put covenant and consolidation merger covenant (see the Appendix for description of each type of covenant). These four types of covenants are considered as the event risk covenants because than can reduce the bond’s exposure to event risks (Cremers, Nair & Wei, 2007). Once the company has one or more of the four types of covenants, the company is coded with a dummy variable of 1. For each characteristics variable, table 2 provides information about number of observations, average, standard deviation, 10th_{percentage and 90}th_percentage.

Since the sample period is limited by the data from TRACE, the sample size in this thesis is quite small. The number of firm-meetings observation is 1119; the number of observations may change for other variables due to missing values.

(17)

4. Methodology

This thesis attempts to examine the effect of corporate governance on bond abnormal return. Most previous literature uses OLS regression with different control variables to estimate this problem, however, the endogenity problem cannot be avoided by using OLS regression. Furthermore, even though there are already a lot of researches in this area, there are no general results of these empirical studies. Based on the study given by Cunat, Gine and Guadalupe (2012), the difficulties of analyzing the relationship between corporate governance and abnormal return may come from two aspects. First, the governance structure of a company generally changes endogenously, which means that the choice of governance structure might be correlated with other firm characteristics. Therefore, regressing the abnormal return with the change of governance structure is more likely to capture the effects of the unobserved firm variables instead of the effects that are estimated. Second, if the market were efficient, all the bondholders would have foreseen the bond prices by using all the information available now. As a consequence, there should be no abnormal return at all. Thus, in order to overcome the difficulties mentioned above, the research design should make sure the governance structure changes exogenously so that the bondholders cannot foresee the price change. Cunat, Gine and Guadalupe (2012) argue that a regression discontinuity combined with an event study can overcome the two difficulties. Therefore, the analysis of this thesis employs the empirical dynamic regression discontinuity model proposed by Cellini, Ferreira, and Rothstein (2010) combined with an event study.

4.1 Regression discontinuity design

Cunat, Gine and Guadalupe (2012) modifies the regression discontinuity model proposed by Cellini, Ferreira, and Rothstein (2010) and combined it with an event study to analyze the impact of corporate governance on stock prices. Since TRACE provides daily bond data as the stock market dose, the model should be practicable in the bond market as well. I use the data from TRACE and the adapted regression discontinuity design for the analysis in this thesis. According to Cunat, Gine and Guadalupe (2012), comparing the bond market reaction to governance proposals between firms with proposals passed by a small margin

(18)

and firms with proposals failed by a small margin can overcome the endogenity problem because the margin is so small that passing a proposal can be seen as an random event, therefore, passing a proposal in this case is not correlated with other firm characteristics. For example, suppose the majority threshold (a proposal will be passed if the vote percentage reaches this number) is 50%, the proposal passes if the vote outcome is 50.1% while the proposal fails if the vote outcome is 49.9%. The 0.2% difference in the vote outcome determines whether the proposal can be passed, in the same time, this difference leads to a discrete change around the threshold in terms of the probability of implementing a proposal. Also, since the vote difference is quit little, the average characteristic with 50.1% of vote and 49.9% of vote might be quit similar, which means other characteristics of the firm other than the probability of implementation are continuous around the threshold. Hence, as long as the probability of implementing a proposal is discrete and the other firm characteristics are continuous around the threshold, the regression discontinuity design can be a solver of the first difficulty that I discussed above. In addition, also because the margin difference is very small, the market is not able to predict whether the proposal will pass, thus, the market cannot foresee the bond price and the expectations of the two types of firm are similar with each other, which means the abnormal returns should be very similar between these two types of companies. The expectations is already incorporated in the bond prices right before the meeting (similar for both types of firms), while the price changes (different for different types of firms) right after the meeting based on whether the proposal passed or not. That overcomes the second difficulty. As a consequence, a regression discontinuity design combined an event study is an appropriate method to estimate the effect of corporate governance change on bond prices.

4.2 Bond abnormal return

Since the bond market is not as liquid as the stock market, the calculation of bond abnormal return can be very problematic. With the introduction of TRACE in 2002, bond prices are available in a daily basis and that makes it much easier to calculate bond abnormal returns. However, not like stock market, TRACE does

(19)

not provide adjusted prices and the result will be biased if I simply use the data of the last trade when the volume of the last trade is small. In order to make the bias as small as possible, I first follow Harris and Piwowar (2006) to delete all trades under $100,000 and then use the price data in the remaining sample to calculate a daily price by weighting each trade by its trade volume. After getting all the daily prices for each bond, I compute the bond returns over the holding period:

(1)

Here I calculate the clean bond return since I do not have access to the accrued interest data. Even though the use of clean bond return may introduce some noises when the accrued interest changes significantly (Bessembinder, Kahle, Maxwell, & Xu, 2009), the bias should be small since I use very small event windows in this study.

In 1984, Hadijnicolaou and Kalay introduce the mean-adjusted model to capture term structure changes. Even since then, this model is always used for bond abnormal return calculation. By using this model, for bond i period t, I first calculate the difference between bond return (BR) and the return on a matched Treasury security (TR) to get the holding period return premium (PBR):

(2) Then I compute the mean premium holding period return (MBR) for the previous T estimation periods:

(3)

Finally, the abnormal return of bond is:

(4) Since some firms have more than one bond exist in the same time, a method is needed to deal with this problem. Previous study either keeps all bonds for each company or chooses one bond as a representative. I use the way that is created by Bessembinder, Kahle, Maxwell and Xu (2009), and they believe this method is more parsimonious than those two methods. They treat the company as a portfolio and the abnormal return of company m is the weighted average of

(20)

the bond abnormal return by trade volume of each bond:

(5) where N is the number of bond issued in a firm and w is the total trade volume of this bond.

4.3 Models

Suppose shareholder in firm m vote share Vmt at an annual meeting day t on a

shareholder proposal. In prevous studies, the very common regression that they use is:

(6) where ymt is the abnormal return of firm m at time t and Dmt is the dummy

variable indicates the proposal passes. captures the effect of corporate governance on abnormal return in firms with passed proposals compared with firms with rejected proposals. However, this regression can be very problematic due to the endogenity problem. Therefore, I use the regression discontinuity model modified by Cunat, Gine and Guadalupe (2012):

(7) where Pl and Pr are different polynomials of order four for observations on the left hand side and right hand side of the threshold respectively. The polynomial is to capture the effect around the discontinuity. are fixed effects for time periods relative to the meeting date, calendar years and focal elections. There is one abnormal return on the meeting day for each firm, while some firms has more than one proposal in each meeting. In order to do the analysis, I need to aggregate the vote come for each meeting. So, equation 7 is modified as:

(8)

where captures the effect of passing a governance proposal on bond abnormal returns, which is exactly what I want to analyze.

(21)

5. Results

Table 3

Pre-meeting Differences between Firms with Passed or Failed Proposals

The table tests whether passing a proposal on the meeting date is systematically related to firm characteristics between firms with proposals passed in the meeting and firms with proposal failed in the meeting. Note that in panel A t refer to days, in panel B and C, t refer to years. Each row corresponds to a different dependent variable and each entry comes from a separate regression. Each entry in the table reports the coefficient on whether a proposal passed. Columns 1 and 2 report the estimated effect of passing a vote on outcome variable levels the year before the annual meeting and the t statistics is shown in parentheses. Column 2 includes the polynomial in the vote share of order four on each side of the threshold. All columns control for year fixed effects and standard errors are clustered at the firm level. Significance at the 1%, 5%, and 10% levels is indicated by ***, **, and *, respectively

Before meeting (t-1) (1) (2) Panel A

Bond abnormal return 1 day before meeting 0.0042 -0.1116 (0.4585) (-0.6134) Bond yield spread 0.0299 -0.0391

(0.8921) (-0.3462) Panel B Tobin's Q -0.1464 0.1970 (-0.0761) (0.1637) Capital expenses/Assets 0.0100 0.0052 (0.0042) (0.0089) Return on equity -0.1049 0.4909 (-0.4598) (0.9956) Panel C G-index 1.1798 -0.0164 (0.1800) (-0.3809) Covenants 0.0507 0.1204 (0.8822) (0.1891) Investment grade -0.0632 -0.0502 (-0.2685) (-0.8412) Polynomial in the vote share no yes

5.1 Pre-analysis test

The main assumption of the regression discontinuity design is that there is no significant difference between the firm characteristics in firms pass with a small margin and that in firms fail with a small margin (Cunat, Gine and Guadalupe,

(22)

2012). In this section, I check whether there exist such differences. Table 3 shows the results of the test. Panel A in column 1 estimates whether firms with passed proposals have different bond abnormal returns and bond yield spread compared with firms with proposal rejected one day before the meeting date. Pane B and C evaluate whether firms with passed proposals have different firm characteristics compared with firms with proposal rejected one year before the meeting date. Column 2 is almost the same as column 1 besides that column 2 controls for polynomials in each side of the threshold to capture the discontinuity. It can be seen from the table that no matter control for the polynomials or not, there are no significant differences between these two types of firm.

5.2 Abnormal returns

Table 4 indicates the effect of governance proposals on abnormal returns in different intervals. The regression estimates the difference of the effect of governance proposals on abnormal returns between firms with passed proposals and firms with rejected proposals on the meeting day. Column 1 to 6 follow equation 6 and column 7 follows equation 8. Column 1 uses the whole sample and since all firms are in this estimation, it is not surprising that there finds no significant differences. Column 2 restricts the sample to the vote-threshold difference is larger than 10% or smaller than -10%, and still there is no significant difference since the sample is still large. Column 3 and 4 restrict the sample to proposals that fall within 10% and 5% of the threshold respectively and the estimates are negative but still not significant. Column 5 and 6 restrict

Table 4

Bond Abnormal Return in Different Bins

This table shows the results of the abnormal returns on the meeting day (t=0). The independent variable is the dummy variable "passed". Column 1 to 6 follow equation 6 and column 7 follows equation 8 with polynomial. Abnormal returns are computed using the mean-adjusted model. Column 1 estimates are based on the whole sample. Column 2 restricts the sample to nonclose votes, that is, votes beyond 10 points of the threshold. Column 3 restricts the sample to

observations with a vote share within 10 points of the threshold, column 4 to five points, and so forth. Significance at the 1%, 5%, and 10% level in indicated by ***, **, and *, respectively.

All Shareholder Proposals

(1) (2) (3) (4) (5) (6) (7) All Votes Nonclose -10;+10 -5;+5 -2;+2 -1;+1 Full Model Pass 0.0039 0.0678 -0.0023 -0.0086 -0.0023** -0.0211*** -0.0211*** (0.1116) (0.2392) (-0.89237) (-0.9753) (-1.8982) (-2.4974) (-2.9471) Observations 2865 1980 892 409 239 104 2865 R-square 0.0003 0.0007 0.0072 0.0014 0.0001 0.0190 0.0110

(23)

Table 5

Abnormal Returns of Passing Governance Proposals

The table indicate the effect of passing a proposal on abnormal returns on the meeting date (t=0) and on the day after (t+1). The dependent variable abnormal returns computed using the mean-adjusted model. All the regressions follow equation 8 with polynomial. Column 1-4 use the full sample and column 5 and 6 uses "G-index" proposals compared with "Other" governance proposals. In column 6, it is a dummy variable for whether abnormal returns are positive or negative. Column 3 uses two dummy variables, covenant and non-investment graded. Column 4 allow for five different dummy variables to capture the number of proposals (one to five) that passed at the meeting. T-statistics are in parentheses. All columns control for year fixed effects, firm-meeting fixed effects; standard errors are clustered by firm. Significance at the 1%, 5%, and 10% levels is indicated by ***, **, and *, respectively.

Abnormal Returns Abnormal Returns G-index versus Other

(1) (2) (3) (4) (5) (6)

G-index Positive return G-index

Day of vote, t -0.0211*** _{Day of vote, t} -0.0129*** -0.0029***

(-2.9471) (-2.8061) (-3.0285)

One day late, t+1 -0.008 -0.008 _{One day late, t+1} -0.007 -0.0632

(-0.9721) (-0.9721) (0.9247) (1.1824)

Day of vote, t Other Other

One vote passed _{-0.0211*** Day of vote, t} -0.0034 -0.0006

(-2.9471) (-0.47519) (-0.8942)

Two vote passed _{-0.0231*** One day late, t+1} -0.0013 -0.0594

(-3.0029) (-0.02841) (-0.3599)

Three vote passed _-0.0329***

(-3.0263)

Four vote passed _-0.0374***

(-3.0392)

Five vote passed _-0.0396***

(-3.6829) Covenants 0.0027*** (2.4901) Non-investment graded 0.0159*** (2.6073) Observations 5595 5595 5595 5595 5595 5595 R-square 0.003 0.004 0.004 0.002 0.005 0.003 Number of firm-meetings 1119 1119 1119 1119 1119 1119

the sample to proposals that fall within 2% and 1% of the threshold respectively. Even though the number of observation is quite small, the estimate is negative and significant. In column 7, I use the full regression model with two polynomials with order for in each side of the threshold and find similar results as in column 6 which means passing shareholder proposals leads to a negative 2.11%

(24)

abnormal return compared with rejecting shareholder proposals.

In addition to table 4 and take into account the fact that a firm can vote for several proposals in one meeting, I use the model in equation 8 in my analysis and introduce several dummy variables in order to test the hypotheses I developed in section 2.3. The results are reported in table 5 and all the columns are controlled for polynomials in each side of the threshold. Column 1 and 2 show the effect of passing governance proposal(s) on abnormal returns on the meeting day and one day after the meeting date respectively. From the table we can see that the effect on abnormal return is only significant on the meeting day. Therefore, the use of the data on the meeting day in table 4 is reliable. The estimate is -0.0211 and significant which is the same as the estimate in table 4 and consist with my first hypothesis. In column 3 I introduce two dummy variables, namely, covenants and non-investment graded to test the second and third hypothesis. The estimate of covenants is 0.0027 and the estimate of non-investment graded is 0.0159 and both of them are significant. It means that passing a proposal in a company with covenants yields a 0.27% abnormal return and in a non-investment graded company yields a 1.59% abnormal return. The results consist with the hypotheses and show that covenants may align shareholders’ and bondholders’ interests and bondholders in non-investment graded firms benefit from shareholder proposals in general. In column 4, I take use of five dummy variables to indicate how many proposals passed in one meeting for each firm. The results report that even there is only one proposal passes, the impact on abnormal return is significant already. And as the number of passed proposals increases, the effect becomes even more negative and all the results are significant. In column 5 and 6 I test hypothesis 4. The difference between column 5 and 6 is that in column 6 I use a dummy variable as the dependent variable. The dummy variable equals to 1 if the abnormal return is positive. According to Cunat, Gine and Guadalupe (2012), even though the dummy variable cannot capture the magnitude of the abnormal return anymore, the use of dummy variable can reduce the bias comes from the outliers. For G-index, the estimates on the meeting day in both column 5 and 6 are negative which is different from my hypothesis. The estimates for Other-type proposal are negative as well but not significant.

(25)

6. Conclusion

This thesis analyzes the effect of corporate governance on bondholder value in S&P 1500 companies during the period 2002-2013. By employing the regression discontinuity design modified by Cunat, Gine and Guadalupe, 2012, I am able to overcome the endogenity problem and other methodology problems existed in previous literature. I find that passing a proposal has negative impact on bond abnormal returns. The use of event risk covenants aligns the interest of bondholders and shareholders so that the impact of passing a proposal on bond abnormal return is positive. The impact is also positive in non-investment graded firms because those firms are under great credit risk and the pass of a proposal improve the corporate governance of the firm and reduce the default risk. Therefore, in general bondholders’ benefits from the improvement of corporate governance in non-investment graded firms compared with investment-graded firms. My analysis also indicates that the more proposals passed the larger the negative impact on abnormal return. All those findings consist with the existing literature and the hypotheses that I made in section 2.3. However, I also find that the G-index proposal has negative impact on abnormal return as well, even though previous studies find that anti-takeover provision can benefit bondholders.

Even though I manage to use the regression discontinuity design to overcome the endogenity problem and are able to find the causal relationships, there are still some limitations in this thesis. First, since TRACE was introduced in 2002, the sample period is very short. Also, in the first three to four years (2002-2005), the data in TRACE is not very complete. All these reasons make the sample size very limited and small. Therefore, the results must be interpreted with caution and it is suggested to do this analysis again when the data is more adequate in the future. Second, due to the differences on regulation conditions and corporate governance polices and culture in different countries, the findings of this analysis may be unique to the US bond market and not applicable to other countries.

It will be interesting to analyze the impact of other five different types of proposals on the bond abnormal return separately when the sample size

(26)

becomes larger in the future. Since in my analysis, only G-index proposal has enough observations for the analysis.

(27)

Appendix

Covenant Description

Covenant type/name Description

Leverage covenant

Negative pledge covenant The issuer cannot issue secured debt unless it secures the current issue on a pari passu basis

Funded debt Restricts issuer from issuing additional funded debt. Funded debt is any debt with an initial maturity of one year or longer

Funded debt subsidiaries Restricts issuer's subsidiaries from issuing additional funded debt) debt with an initial maturity of longer than one year)

Stock issuance Restricts issuer from issuing additional _{common stock}

Stock issuance subsidiaries

Restricts issuer from issuing additional common stock in restricted subsidiaries. Restricted subsidiaries are those which are considered to be consolidated for financial test purposes

Subordinated debt issuance Restricts issuance of junior or subordinated _debt. Borrowing restricted Subsidiaries are restricted from borrowing, _{except from parent} Networth covenant

Maintenance networth Issuer must maintain a minimum specified _{net worth.} Indebtness Restricts user from incurring additional debt with limits on absolute dollar amount of debt

outstanding or percentage total capital Indebtness subsidiaries Restricts the total indebtedness of the _{subsidiaries.} Leverage test Restricts total-indebtedness of the issuer Leverage test subsidiaries Limits subsidiaries' leverage

Poison put covenant Upon a change of control in the issuer, bondholders have the option of selling the issue back to the issuer (poison put). Consolidation merger covenant A consolidated merger of the issuer with _{another entity is restricted.}

(28)

Reference

Anderson, R. C., Mansi, S. A., & Reeb, D. M. (2003). Founding family ownership and the agency cost of debt. Journal of Financial economics, 68(2), 263-285.

Bessembinder, H., Kahle, K. M., Maxwell, W. F., & Xu, D. (2009). Measuring abnormal bond performance. Review of Financial Studies, 22(10), 4219-4258.

Bhojraj, S., & Sengupta, P. (2003, July). Effect of corporate governance on bond ratings and yields: The role of institutional investors and outside directors. The Journal of

Business, 76(3), 455-475.

Black, B. S., Jang, H., & Kim, W. (2006). Does corporate governance predict firms' market values? Evidence from Korea. Journal of Law, Economics, and Organization, 22(2), 366-413.

Cellini, S. R., Ferreira, F., & Rothstein, J. (2010). The value of school of facility investments: Evidence from a dynamic regression discontinuity design. The Quarterly Journal of

Economics, 125(1), 215-261.

Coase, R. H. (1937). The nature of the firm. economica, 4(16), 386-405.

Cremers, K. J., & Nair, V. B. (2005). Governance mechanisms and equity prices. The Journal of

Finance, 60(6), 2859-2894.

Cremers, K. M., Nair, V. B., & Wei, C. (2007). Governance mechanisms and bond prices. Review

of Financial Studies, 20(5), 1359-1388.

Cuñat, V., Gine, M., & Guadalupe, M. (2012). The vote is cast: the effect of corporate governance on shareholder value. The journal of finance, 67(5), 1943-1977.

De Franco, G., Vasvari, F. P., & Wittenberg-Moerman, R. (2009). The informational role of bond analysts. Journal of Accounting Research, 47(5), 1201-1248.

DeFond, M. L., & Jiambalvo, J. (1994). Debt covenant violation and manipulation of accruals.

Journal of accounting and economics, 17(1), 145-176.

Drobetz, W., Schillhofer, A., & Zimmermann, H. (2004). Corporate governance and expected stock returns: Evidence from Germany. European Financial Management, 10(2), 267-293.

Ederington, L., Guan, W., & Yang, Z. (2012). Bond market event study methodology. Available

at SSRN.

Ertmur, Y., Ferri, F., & Stubben, S. R. (2010). Board of directors' responsiveness to shareholders: Evidence from shareholder proposals. Journal of Corporate Finance,

16(1), 53-72.

Gillan, S., & Starks, L. T. (1999). Relationship investing and shareholder activism by institutional investors: the wealth effects of corporate governance related proposals. Gompers, P., Ishii, J., & Metrick, A. (2001). Corporate governance and equity prices. National

bureau of economic research, w8449.

Handjinicolaou, G., & Kalay, A. (1984). Wealth redistributions or changes in firm value: an analysis of returns to bondholders and stockholders around dividnd anouncements.

Journal of Financial Economics, 13, 35-63.

Harris, L. E., & Piwowar, M. S. (2006). Secondary trading costs in the munipal bond market.

The Journal of Finance, 61(3), 1361-1397.

Jensen, M. C., & Mecking, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of financial economics, 3(4), 305-360.

Klock, M. S., Mansi, S. A., & Maxwell, W. F. (2005, December). Does corporate governance matter to bondholders? Journal of financial and quantitiative analysis, 40(4), 693-719.

Masulis, R. W., Wang, C., & Xie, F. (2007). Corporate governance and acquire returns. The

Journal of Finance, 62(4), 1851-1889.

Shleifer, A., & Vishny, R. W. (1997, June). A survey of corporate governance. The journal of

finance, 52(2), 737-783.

Sloan, R. G. (2001). Financial accounting and corporate governance: a discussion. Journal of

Accounting and Economics, 32(1), 335-347.

The impact of corporate governance on bondholder value

University of Amsterdam

The Impact of Corporate Governance on

Bondholder Value

Table of Contents