Corporate Governance and the Price of
Debt: an Empirical Study.
Shechter Alon
MSc. Business and Economics, Finance Track Master Thesis
Amsterdam Business School University of Amsterdam
Roeterstraat 11
1018WB Amsterdam, The Netherlands
22th of September 2013
2
Abstract
The debt agency problem and agency cost may suggest that the quality of firm's corporate governance may have an effect on its price of debt and consequently the amount of debt it has on its balance sheet. Using the Sarbanes and Oxley act as a proxy for the change in the quality in corporate governance, we estimate the effect this have
on bond spread and debt to capital ratio. We do this by using a differences in differences methodology, comparing US firms which have adapted the SOX act with non cross-listed European firm which have not. We find a statistically significant reduction in
US firms bond spreads after adopting SOX, suggesting that improved corporate governance can indeed reduces the price of debt.
3 Table of Contents
1. INTRODUCTION ... 4
1. LITERATURE BACKGROUND ... 7
1.1. FIRM CAPITAL STRUCTURE AND AGENCY COST OF DEBT. ... 7
1.2. AGENCY PROBLEM. ... 9
1.3. DEBT AND THE AGENCY PROBLEM. ... 10
1.4. AGENCY COSTS OF DEBT ... 11
1.5. CORPORATE GOVERNANCE AND THE COST OF DEBT ... 15
1.6. SOX AFFECT ON COST OF DEBT ... 18
2. METHODOLOGY... 21
4. DATA SAMPLE ... 25
5. EMPIRICAL FINDING AND RESULTS ... 28
6. CONSISTENCY CHECKS. ... 32
7. CONCLUSION AND FINAL REMARKS ... 34
8. BIBLIOGRAPHY ... 37
9. APPENDIX ... 40
9.1. TABLE 1 : BREAKDOWN OF RELEVANT EMPIRICAL LITERATURE. ... 40
9.2. TABLE 2: DESCRIPTIVE STATISTICS U.S. FIRMS SAMPLE ... 42
9.3. TABLE 3: DESCRIPTIVE STATISTICS EUROPE FIRMS SAMPLE ... 43
9.4. TABLE 4: DESCRIPTIVE STATISTICS ENTIRE SAMPLE ... 44
9.5. TABLE 5: BREAKDOWN OF SAMPLE TO INDUSTRY TYPES. ... 44
9.6. TABLE 6: 2004 REGRESSION VARIABLES CORRELATION MATRIX ... 45
9.7. TABLE 7: 2006 REGRESSION VARIABLES CORRELATION MATRIX ... 45
9.8. TABLE 8: 2004 DIFFERENCES IN DIFFERENCES ESTIMATION OF THE EFFECT OF SOX ON FIRM'S SPREAD AND DEBT TO EQUITY RATIO. ... 46
9.9. TABLE 9: 2006 DIFFERENCES IN DIFFERENCES ESTIMATION OF THE EFFECT OF SOX ON FIRM'S SPREAD AND DEBT TO EQUITY RATIO ... 47
9.10. TABLE 10: 2004 DIFFERENCES IN DIFFERENCES ESTIMATION OF THE EFFECT OF SOX ON FIRM'S SPREAD AND DEBT TO EQUITY RATIO IN EUROPEAN CROSS LISTED FIRM ... 48
9.11. TABLE 11: FIXED EFFECTS PANEL DATA REGRESSION OF US SAMPLE IN THE YEARS 2000-2007 .... 49
4
1. Introduction
One of the goals of Modern corporate finance research is to find out what determines firm's finance structure. The first comprehensive model was envisioned by Modigliani and Miller (1958). In their paper they argued that financing decisions and corporate finance structure are in fact irrelevant, and do not affect the value of the firm as a whole. However this is only true in perfect markets, with no government taxes and no information asymmetries. These inefficiencies tend to distort this conclusion. The fact that corporations invest a great deal of resources to manage their finance is a testament to this.
Corporate governance research is closely related to corporate finance research. Its main goals are to determine the best methods to govern a firm, from the point of view of increasing stakeholder's wealth. The need for a corporate governance system originates from the separation of ownership and control between the investors, who own the company, and the management who controls it (Shleifer and Vishny (1997) page 4). This separation in ownership and control is the root cause of what is called the "agency conflict" between the shareholders and managers, and between shareholders and creditors. Because the managers do not always act in the best interest of the shareholders, they can, in various methods, expropriate the shareholders. A similar problem can occur between shareholders and creditors. Shareholders, who have
indirect control of the firm, through the board of directors, can expropriate the creditors in 2 ways: by transferring the risk of investing in risky assets to bondholders, in what is called "asset substitution" (Jensen and Meckling(1976)), or by failing completely to invest in assets that will serve to pay back the debt , in what is called "underinvestment problem"(S.C. Myers (1977)).
In a world with rational investors, it is assumed that whoever creates these problems bears its costs. If managers expropriate the shareholders, new shareholders who are aware of this will be willing to pay less for the company shares, thereby increasing its cost of capital. The same happens if shareholders expropriate the debt
5 holders- the price the creditors will be willing to pay for company bonds will decrease, resulting in a higher yield and a higher cost of debt.
Good corporate governance may mitigate these agency problems. Extensive research has been done in order to answer this question regarding the price of equity( Sakife, Hollis, Collins, Knney jr Lfond (2009) Gomprs Ishii Metrick (2003), Cremers and Nair(2005), Giroud and Mueller (2011)). However regarding the price of debt there is less research. This may have to do with the Anglo-Saxon view of the firm- that its main goal is to increase the wealth of shareholders, and not the wealth of all its stakeholders. Nevertheless, this does not diminish the importance of such research. If it can be
empirically proven that indeed good corporate governance decreases the cost of debt, then this will motivate firms to improve their corporate governance, in order to lower their cost of debt. This in turn will allow them easier access to funding, improve their growth prospects, and by minimizing inefficiencies improve the economy as a whole.
This paper tries to answer exactly this question, by empirically testing whether corporate governance has any effect on firm's price of debt. Previous research either employed a simple OLS regression methodology ((Amir Yangling and Gilad(2010)), or only tested a closely related subject (Arping and Sautner (2010)).This paper adds to this research by estimating exactly the difference a change in the quality of corporate governance has on the price of debt. We assume that good corporate governance can mitigate the debt agency conflict and minimize the underinvestment problem, and by doing so reduce the firm's cost of debt. Consequently, those firms will have a better access to funding, because they can issue debt at a greater discount, and therefor they will have a higher debt to capital ratio.
To test this assumption we will employ a "differences in differences" estimation method. This method is based on comparing an experiment, either constructed or natural, on both a treatment and a control groups. By analyzing the results of the
experiment we can quantify the effect corporate governance has on the price of debt . We will use the Sarbanes and Oxley act of July 2002 as a natural experiment, and then
6 check its effect on the price of debt on the treatment group, which will be a sample of US based firms, and a control group, which will be a sample of European firms. The SOX act introduced a number of improvements to firm's corporate governance and to the quality of its financial statements. Therefore it is a good indicator for the quality of firm's corporate governance. Furthermore the adoption of SOX might also have an implication on firm's leverage ratio; we will test this assumption empirically .
The rest of the paper is structured as follows: Section 2 is a review of relevant literature and past research. Section 3 defines the methodology that will be employed. Section 4 describes the data sample and includes descriptive statistics. Section 5
presents the empirical findings and results. Section 6 presents consistency tests. Section 7 are conclusions. Section 8 are the references list and Section 9 is an appendix
7
1. Literature Background
1.1. Firm capital structure and agency cost of debt.
What is the nature of the relationship between corporate governance and corporate finance structure? This a highly debated topic, and was researched
extensively in the past ((Rafael, Lopez-de-silanes-Sheleifer, Vishny (2000), Jensen and Meckling (1976)). This is still a relevant topic today.
In their ground-breaking article ModiglianiMiller (1958), showed that in a perfect capital markets, capital structure has no effect on firm’s value. That is because every finance decision the firm makes can be done by individual investors, therefore in a non arbitrage scenario no added value can be created. The set of rules which defines a perfect capital market are: Investors and firms can trade the same set of securities at a competitive market; there are no taxes, transaction costs, or issuance costs; Firms finance decisions do not change future cash flows generated by its assets (Fama and Miller 1972).However in reality, firms invest significant amount of time and money in managing their capital structure. This is due to several factors, the most important one being the existence of corporate tax. In most countries firms can deduct interest
expenses created by debt from their earnings, thus resulting in less tax payments for the firm. This gain for the companies is called tax shield. The present value of all of these tax shields essentially can be viewed as a new asset, which is not listed in the balance sheet, but which increases the economic value of the firms. This new asset is essentially the corporate tax rate *debt (assuming no changes to corporate tax rate) (Modigliani
andMiller (1963)). It is a linear function increasing in debt, up to the point where the tax shield equals fully to the earnings. Knowing this, Companies will prefer to fully exploit this tax shield, as it maximizes their value. However in reality, this is not the case. Most firms chose to shield only about 40% of their taxable income (Kemsley and Nissim
8 (2002)), resulting in suboptimal capital structure. Clearly there are other factors at play here. these are bankruptcy, and agency costs .
Bankruptcy costs are the costs incurred by a firm who goes bankrupt. These costs can be separated into direct costs and indirect costs. Direct costs are costs associated with bankruptcy process itself: lawyers and accountants fees, other
professional fees, etc. Indirect costs are costs which are not related to the bankruptcy process itself, but which the firm incurs because of the bankruptcy. This may include: loss of good will, worse credit conditions, loss of customers and suppliers, etc. These costs downgrade the value of the firm, even before the event of a bankruptcy itself, because of the risk of this bankruptcy to happen. In a nutshell, it is the loss of opportunity the firm incurs. This also is highly dependent on market view on the bankruptcy itself. customers expecting a bankruptcy might stop buying the firm products, fearing it will not support them and its market value will drop. On the other hand it the bankruptcy can be perceived by customers as only a formal procedure, and its market value will therefore not drop. Rational creditors are aware of these direct and indirect costs, and in highly leveraged firms ((roughly over 50% of debt to total firm ratio) Hayne and E. Klaus Bjerre (1996)), they will subtract the sum of these costs from the total credit they supply to the firm, resulting in effective higher cost of debt. Gruber and Warner (1977) have found empirical evidence for such direct, and indirect costs. They compiled a list of 11 American railroad companies who went bankrupt. Then they calculated the median direct bankruptcy costs to be 2 million. This amount is rather small, compared to the average firm market value in the sample of 250 million (before the bankruptcy) . As a proxy for the indirect costs they measured the median loss in market value in an event of a bankruptcy, seven, five, two and one year before it
happened. They calculated the median market value loss in firms seven years before the bankruptcy event to be 78.9%. Clearly this is a sizable amount, and in large firms it can amount to sums much larger than the direct bankruptcy costs. Therefore it seems that the indirect bankruptcy costs are much more substantial and important than the direct costs when determining the cost of debt.
9
1.2. Agency problem.
The Second factor which affects the price of debt is agency costs. Agency costs are caused by equity or debt. They increase marginally, up to a point in which it is less costly for a firm to be financed by equity than debt. In other words, in equilibrium a firm will be financed by equity and debt because of this fact.
The agency problem is the root cause of these costs, both in outside equity and debt. Jensen and Meckling (1976), explain that the agency problem arises from the separation of ownership and control in publicly traded firms. An agency relationship is defined as a “contract in which one or more person engage another person to perform service on the behalf of the former" (Jensen and Meckling (1976) page 4)). In corporate finance, the management is the agent of the shareholders (and creditors). Both the agent and principle aim to maximize their welfare. However their utility functions are not always the same. Therefore the management does not always work in the best interest of the shareholders. One example of this is the so called "empire building" done by management. When a firm has surplus cash deposits generated from free cash flows, the management may chose to invest these funds in low net present value projects, rather than paying it out as dividends or share repurchase to the investors, increasing the firm size beyond its optimal size. This is because the larger the firm is, the bigger the power the management has, by increasing the resources they manage. Also manager's compensation is correlated with firm size. (Jensen 1986). In order to prevent such behaviour and to insure the return on their investment, shareholders must monitor the agent, by establishing both monitoring measures (financial disclosures for example), and by expending resources to incentivise the management to work in their best
interests(bonding costs).A good example for this is executive compensation package consisting of both a monthly salary and stock options. Another example is bonus granting based on performance.
11 Agency costs therefore is the sum of 3 different things: the decrease in principle wealth due to agent and principle divergence of interests, the costs of the monitoring measures applied by the principle to mitigate this problem (financial statements for example), and bonding costs.
1.3. Debt and the agency problem.
Issuing debt to creditors debt is one method in which a firm can mitigate the shareholders management agency problem, and improve its corporate governance. First hypothesized by Jensen (1986) ,debt can reduce the agency problem created by surplus cash flows by binding the promise to pay out these cash flow to the stakeholders, rather than spending them on low value assets, or empire building. Empire building , as we recall, is a process in which the management entrench itself, by increasing the size of the firm beyond its optimal size. By doing so they destroy value to the shareholders, which could have been paid out to shareholders either in the form of dividends or share repurchase. If a management believes that the increase in cash flow is permanent, or would be permanent in the foreseeable future, it may announce a permanent increase in dividends. However such a promise is weak in the sense that it is not a legal
obligation, and therefore these dividends can be reduced in the future. The markets perceive this is a such, and in the event of dividend cut the price of the share drops substantially( Aharony and Swary (1980)).
Debt principle payments ,however, are a legal obligation backed by the threat of bankruptcy. Debt therefore can be used to bind future promise of the management of paying out cash flow to shareholders, by issuing debt in exchange for stocks. If the firm fails to keep on its promise the creditors (former shareholders) can initiate a bankruptcy procedure. In this sense Debt can be a substitute for dividends , and is a control
mechanism of the stakeholders on the management. Debt issuance, however does not always have a positive effect on firm value. beyond a leverage ratio of 50% the firm
11 incurs bankruptcy costs. Also high level of debt can result in another kind of agency conflict - an agency conflict between the shareholder and the creditors, known as "debt agency problem". This too has an effect on firm capital structure, and more precisely the cost of debt.
1.4. Agency costs of debt
The debt agency problem exists in highly leveraged firms, where manager's main concern is to maximize shareholders wealth, rather than whole firm value (Jensen and Meckling (1976)). This happens when creditors and equity holders are 2 different sets of investors. When this occurs the management may chose to transfer wealth from
bondholders to equity holders, by the process of "risk shifting" (or "asset
substitution"(Jensen and Meckling (1976))), or doing what is called "underinvestment problem" (Myers 1977)
"Risk shifting", or "assets substitution", is the transfer of wealth from creditors to shareholders, by increasing the risk of the firm's assets. To understand this think of the next example: Firm A is financed mostly with debt, D, and a small amount of equity, E. There are 2 time dates: 0 and 1. The company can invest in 2 different projects which have the same expected payoff E[X], but which one is riskier than the other: . In this case we can view the equity as a call option on the value of the firm V, with the exercise price being the total amount of debt, or D. This is because in a case of a default, the equity holders lose only their initial investment and can essiently walk away. this is similar to not exercising the option. Black Sholes model for valuating options tells us that the value of a call option increases with higher volatility, or risk. Knowing this, managers who work in the best interest of the shareholders will tend to invest in riskier projects, by substitute safe assets with risky assets, thereby increasing volatility of the income for the whole firm and the value of the equity, even up to the point of choosing risky negative N.P.V projects, which may decrease the value of the firm as a whole. This
12 increase in equity value is actually at the expense of the bondholders, and constitutes a wealth transfer from bondholders to equity holders. This exemplifies the fact that in highly levered firm equity downside risk is limited, while bondholder's risk is not (See also John and Senbet (1998)). The management tends to do this "risk shifting" ex-post, after the debt contracts are in place. Creditors with rational expectations are well aware of this "risk shifting problem", and will be willing to pay less for the bond contracts, thereby increasing the firm's cost of debt, and also transferring the risk back to the equity holders. The reduction in price will be equivalent to the aforementioned wealth transfer from the bondholders to the equity holders. Therefore the debt agency costs are exactly the increase in the price of debt caused by this "risk shifting.
There is growing empirical literature for this risk shifting behavior. Eisdorfer (2008) empirically tested the effect of market uncertainty on investment intensity in healthy firms and financially distressed firm. His aim was to empirically test whether the "risk shifting" agency problem and its costs really exists. His hypothesis was that times of market uncertainty have a positive effect on investment intensity in financially distressed firms. He also hypothesized this uncertainty has a negative effect on the value of the assets being invested by financially distressed firms, compared to healthy firms. To achieve this first he estimated market volatility using GARCH time series model. He defined investment intensity as the gross capital expenditure divided by PP&E*1. To measure the level of financial distress in a firm he used Altman’s Z score2. He then estimated, using a panel data OLS method, the effect of market volatility on
investment intensity in healthy firms and financially distressed firms. He found statistically significant negative effect of volatility on investment intensity in healthy firms, and significant positive effect on investment intensity in financial distressed firms, validating his first hypothesis. Next he estimated investment intensity effect on firms return on equity, in healthy firm and financially distressed firms, separating between
1
property, plant, and equipment Fazzari, Hubbard, and Petersen (1988)
2 Z-score = 1.2(Working capital/Total assets)+ 1.4(Retained earnings/Total assets)+3.3(Earnings before
interest and taxes/Total assets)+0.6(Market value of equity/Book value of total liabilities) + 0.999(Sales/Total assets).
13 times of high market volatility and low market volatility. Using a panel data
methodology , his estimations did not suffer from omitted variable bias. problems. Again he found that the link between investment intensity and return on equity is much less significant in financially distressed firms, validating his second hypothesis. While his article only links between market volatility and investment in financially distressed firms, and not a direct link between amount of debt and risk shifting , it does show that in times of uncertainty financially distressed firms tend to invest more ,and in lower value assets, then healthy firms. In other words shareholders in financially distressed firm exploit market volatility to do assets substitution, presumably in order to increase the value of their shares.
Another aspect of agency costs, which is different to the assets substitution problem, is the debt underinvestment problem. C. Mayers (1977) in his article describes in detail this behavior. When firms are heavily levered, they might not only choose to invest in risky assets instead of safer assets, but also pass on riskless, positive value investment opportunities all together. This may happen when the debt matures after the investment opportunity has expired, the new asset requires additional investment by shareholders ,and there is a risk the company would not be able to pay back its debt. When debt repayment is high enough, shareholders investment is also high and the value of the new asset is sufficiently low(but still positive), CEO who maximizes the shareholders welfare might pass on safe investments. This is because the entire new value created goes as repayment to debt holders, while shareholders will not benefit from investing at all. In essence the shareholders new investment will serve to pay back the debt to the debt holders.
Who bears these agency costs then? Is it the debt holders, or the shareholders? Similarly to shareholders management agency problem, In a world with rational
investors, it is assumed that the shareholders bear these costs. Creditors are fully aware of the risk shifting and underinvestment problems, which in turn increase the riskiness of the debt itself and the volatility of the expected payment. They therefore require a
14 higher cost of debt, in the form of bigger spreads, and lower bond price. therefore firms can raise fewer funds from debt.
To what extent can a firm decrease the agency cost of debt, and reduce its cost of debt? There are 2 important mechanisms to achieve this: convertible bonds and improved corporate governance.
A convertible bond is a kind of a debt contract with the option to convert it to a predetermined number of shares in what is called "conversion ratio". Convertible bonds can be thought of as a combination of a straight debt, and a call option written by the firm on its own shares, called a warrant. When the warrant is exercised, the firm issue new shares to debt holders, in exchange of the debt itself, with the number of shares determined by the conversion ratio. Green (1984) ,in his article, has found that by using a conversion warrant in a straight bond, firms can change the payoff function to equity holders from risky investments to a be concave function, from a convex function. This has the effect of reducing shareholders upside gains from taking risky investments, because owners of convertible debt will also share them. This in turn will decrease the shareholders incentive to invest in risky assets, and help to decrease the agency conflict between the shareholders and creditors, even without exercising the option itself. Furthermore, any increase in assets risk will not affect convertible bond value. That is because the decrease in the value of debt due to agency costs is offset by the increase in value of the conversion option. So in essence a convertible bond helps to align the interests of the shareholders and equity holders.
However Convertible bonds can only solve the agency debt problem partially. This due to 2 reasons. first convertible bonds are offered at a discount because of the conversion option. Second, and more importantly, convertible bonds are mostly a tool of the management for delay equity issuing. Jermey c. (1992) has in his article has developed a model explaining that major reason for issuing convertible bonds is information asymmetry. firm's may use convertible bonds when they wish to issue equity, when their equity prices are low due to information asymmetry. To be successful
15 in this firms must be sure their equity price will rise in the future, or include a callable option. If those bonds will not be converted on the future, they may lead to additional costs of financial distress. It is trying to achieve 2 goals with a single instrument, and is the main reason why it cannot solve the agency problem completely.
1.5.
Corporate governance and the cost of debt
Shleifier and Vishny (1997) in their paper provide a survey of corporate
governance. They define corporate governance as “ways in which suppliers of finance to corporations assure themselves of getting a return on their investment” (Shleifier and Vishny (1997) page 1). In other words, it is the different measures designed to solve, albeit partially, the agency problem arising from the separation of ownership and control in public firms. The extent in which these measures help solve the debt agency problem and reduce firms cost of debt, by minimizing aforementioned agency costs, is a matter of great importance to both the companies themselves and investors. A small number of articles have estimated this. Table 1 in appendix 9.1 provides a breakdown of relevant empirical literature.
Anderson Mansi and Reeb (2004) check one measure of corporate governance on the price of debt, namely the board of directors. One of the main responsibilities of the board of directors is monitoring the management, thereby insuring creditors (and shareholders) that they are not expropriating them. The authors checked whether different characteristics of the board effect firm’s cost of debt: board independence, size of the board, audit committee independence and size, board members financial expertise and frequency of audit committee meetings. They did this by estimating how these factors effect firm credit rating and bond yield spread, in a panel data regression with 1052 observations. They found that board
16 and frequent audit committee meeting reduces firms bond yield spread, thereby reducing its cost of capital. However they did not find any effect of board members financial expertise on bond spread. It seems that creditors do not consider this important in their decision making. Because their method of estimation was panel data regression their results are robust and do not suffer from omitted variable bias.
Bradley, Chen, Dallas and Snyderwine (2008) have researched this more extensively, by estimating the effect of various corporate governance measures on firm's bond spread, and credit ratings. These corporate variables are separated into: ownership structure, earnings quality, whether incorporated in Delaware, the Board structure and executives compensation. They also included in the regression the
GINDEX3, an index that measures corporate governance. In both models they have
used an unbalanced panel data of 775 observations, taken from COMPUSTAT database and S&P property database, in the years 2001-2007. In the first model they employed a logit regression, with firm financial controls, to estimate the effect of the
aforementioned variable, on firm's credit ratings. Like Anderson Mansi and Reeb(2004) they have found statistically significant link between board structure and credit
ratings, namely that board size increases credit rating while the percentage of directors with zero equity decreases credit rating. Presumably directors with equity holding are perceived by creditors to be more aligned with firm's goals. They also found that ownership structure has a negative effect on credit rating: The number of insider directors and number of share block- holders decrease credit ratings. Also executive total bonus has a negative effect on credit rating. Next they estimated the effect of corporate governance on bond spread. They did this by using a similar model, this time with the dependent variable being bond spread. In this estimation they got somewhat different results: Interestingly, in contrast to the affect on credit rating board structure has no statistically significant effect on bond spread. However CEO base salary, which is defined as annual base salary of the CEO as a percentage of total compensation, increases bond spread by 1.02%. This is a different then the result they
3
“Governance Index” constructed by Gompers, Ishi and Metrick in “Corporate Governance and Equity Prices,” (2113)
17 came up in the previous estimation(no link between CEO pay and credit rating). It seems that Bond traders are concerned with CEO payment when it comes to bond spreads. However the rating agencies does not take this into account when
determining their ratings. All other results are consistent with previous estimation. Klock, Mansi and Maxwell (2005) came up with similar results when they estimated a similar model in a panel data with 1877 observations spanning the years 1990-2000, using a different databases- The Lehman Brothers Fixed Income Database and the Investor Responsibility Research Centre. They too found significant negative effect of both the Gindex and the percentage of institutional investors on bond yield spread.
Bhojraj, Sanjeev and Partha Sengupta (2003) in their paper came up with similar results. In their article they also checked the affect the interaction between bond rating and institutional investors and outside director had on bond yields, again employing panel data regression with 2098 observations of the years 1991-1996. They found significant negative effect of bond rating and institutional investors on bond yields, but not the same effect for bond rating and outside directors.
However it may be that intuitional investors prefer to invest in less risky, lower yield bonds, then higher yield, more risky bonds, meaning that they have no effect on the cost of debt. This is a reversed casualty problem , which in turn is a part of the more generalized endogeniety problem. Endogeneity4 is defined as a correlation between one or more of the independent variable and the error term. Not controlling for endogeneity may result in a bias estimation. This problem occurs either because of omitted variable bias, simultaneous equations, or autocorrelation. Solving a this problem is tricky; while omitted variable bias is easily solved by using a panel data or diff in diff method, it is not always easy to spot a simultaneous equation problem or reversed causality. In such a case the best solution is to perform an instrumental variable regression. The authors did just that.To exclude an such endogeneity problem between bond yields and institutional investment, they implement a simultaneous
4
Historically coined to describe a variable which is determined within the model itself (endogenous), as opposed to exogenous variable ,which is determined outside of the model
18 equation approach and estimated this using a 3 SLS5 method. They have found that indeed institutional investors prefer to invest in low risk low yield bonds; however there is still statistically significant effect of institutional investment on bond yields.
All these articles are evidence to the link between corporate governance and the price of debt. However all of them were employing a an OLS panel data regression method, which may not capture fully the effect of corporate governance has on the price of debt. There might be different, unspecified factors, which are linked to
corporate governance and reduce the cost of debt. To solve this problem we must look to an exogenous shock, which improved corporate governance, and has no causal link to other factors. Indeed there is such a shock, the introduction of the Sarbanes and Oxley act in July 2002.
1.6. SOX affect on cost of debt
The Sarbanes and Oxley act of 2002 was enacted in reaction to a number of corporations and accounting scandals, most of them caused as a result of bad corporate governance, and fraudulent financial reports. This act was an attempt to prevent such occurrences from happening again, by improving corporate governance. It has introduced 11 new rules that companies must adopt, including: establishment of PCAOB, a non-profit corporation which oversees the audits of public companies, a further increase in auditor independence, introducing executive individual
responsibility for the accuracy of financial reports, enhanced financial disclosures, employing measures to minimize analyst conflict of interest, etc. The SOX act also defines the authority of the SEC, enacts certain criminal penalties for corporate fraud, increases the penalties in white collar crimes, defines "corporate fraud" as criminal offends and increases the penalties for committing these frauds. All of these measures aim is to decrease the probability of corporate fraud, and increase corporate
5
19 governance monitoring on management. Therefore they are good proxy for the
change in the quality of corporate governance.
Number of articles empirically examined if SOX indeed had the desired result. Andrade, Sandro, Gennaro, and Hood (2009) checked the effect SOX had on the credibility of financial reporting. Using information on firms CDS, they constructed a "corporate opacity" variable. A CDS, or credit default swap, is an over the counter insurance contract on debt. CDS and corporate bond yields are closely related; however CDS are more liquid, simplifying the gathering of data. Using CreditGrades CDS pricing model6 they calculate the model implied CDS spread. They then defined corporate opacity as the difference between the total debt on the balance sheet, to the amount of debt that will drive a corporation to a default. They calculate this using the difference between market base CDS and the model implied CDS spread. They then subtracted the mean and median corporate opacity of firms before implementation SOX and after implementation of SOX. They find statistically significant difference of -0.206 and -0.177 in corporate opacity respectively. This means that SOX had slightly increased the
credibility of financial reporting. However this in fact does not explain whether SOX had any influence on the cost of debt itself. Also, by employing simple OLS methodology, their estimations may suffer from omitted variable bias, which undermines their results.
A more relevant research is the paper by is Amir Guan and Livne (2009).In it they tried to indirectly estimate the effect the adoption of SOX had on firm's credit rating and spread. One of SOX aims was to increase auditor independence. they hypotheses that doing so will also lower the cost of capital. They found that indeed auditor
independence lowers bond spread and increase bond rating, and also that SOX increased auditor independence. However they did not empirically examined a direct link between SOX and bond rating and spreads, only an indirect effect. Also , similarly to Andrade, Sandro, Gennaro, and Hood (2009), they have used in their methodology
6
Model was jointly developed by Goldman Sachs, JP Morgan and Deutsche Bank and is used by practitioners.
21 simple OLS regression, controlled for changes in firm characteristics. This is not
econometrically viable methodology, as it susceptible to omitted variable bias problems, which might hamper their results.
The last and most relevant article to this topic is by Arping and Sautner (2010). In it they measured the effect SOX had on firm's opaqueness, by using a similar
differences and differences methodology. To proxy the level of opaqueness they opted to use analyst's forecasts. They assume that SOX decreased opaqueness, and therefore analyst's forecasts will turn out to be more accurate. To measure firm opaqueness they constructed 2 variables: Forecast error and Forecast Dispersion. Forecast Error is the difference between the mean analyst forecast error of firms earning per share and the actual data, scaled by actual data. Forecast Dispersion is the absolute standard deviation of EPS forecast divided by the mean forecast. Next they compared a sample of European firm, which are cross listed and therefore must comply with SOX with a control group of European firms, which are not cross listed, and therefore do not comply with SOX. They employed a diff and diff approach, with the time period before SOX being the years 2001-2004, and the time period after SOX being the years 2005-2007. They found that indeed there is a significant decrease in forecast error and forecast dispersion for both cross listed and non – cross listed firms. However the decrease is much bigger for cross listed firms – a 37% larger reduction in forecast error, meaning the European firms, which have adopted SOX, are less opaque to analysts. They concluded that indeed the SOX act helped to reduce firm opaqueness, and improve its financial transparency.
21
2. Methodology
The previous papers by Anderson Mansi and Reeb (2004), Bradley, Chen, Dallas and snyderwine (2008), Klock, Mansi and Maxwell (2005) and Bhojraj, Sanjeev and Partha Sengupta (2003) all showed a plausible decrease in the price of debt due to improved corporate governance. Building on this we argue that SOX also reduces the cost of debt by improving corporate governance. we assume that better corporate governance leads to lower cost of debt, by mitigating agency cots. This will be
empirically tested, using a similar approach to Arping and Sautner, with modifications. Our hypothesis is then as follows:
1. : SOX has no statistically significant effect on the cost of debt
2. : SOX has reduced the cost of debt in U.S. firms, compared to European firms. In the previous papers mentioned above researchers have opted to use an OLS method to estimate this. However this methodology is susceptible to omitted variable bias problems which may undermine the results. For this reason this paper will employ a differences in differences model regression method. The differences in differences regression model is unique in the sense that it allows us to precisely measure the effect the change in corporate governance had on the cost of debt. This change will be The SOX act, which had improved corporate governance, through various measures, most notably in increased auditor independence and better financial reporting. Therefore the adoption of Sarbanes and Oxley act will constitute as the exogenous shock. Using the diff and diff regression model, controlled for financial factors, eliminates any
endogeneity problems that may arise from omitted variable bias ,that may relate to implementation of SOX , and which have an indirect effect in debt. This is due to how the regression itself is structured. The diff in diff estimator compares the average change a variable had on both the control and treatment groups over a selected time period. any variable which is not specified in the model is eliminated, resulting in unbiased estimation as follows:
22
This methodology hinges on the fact that change is different between the control and treatment groups, while the common trend is similar,. to make sure the trend is similar, we include financial controls.
To estimate this in a cross section diff and diff regression, we will use a list of US Based companies traded on the NYSE and NASDAQ, as the treatment group. This is the main difference between this method and the method employed by Arping and Sautner (2010). Unlike Arping and Sautner (2010) , we opt to compare US firms to non-cross listed European firms, rather than comparing cross listed European firms to no cross listed European firms. This is due to availability of data - there is much more data on US bonds the their European counterparts. Indeed European and Us firms differ in
corporate governance. However what we are checking is the effect of a single corporate governance reform on the price of debt. Astrid van Landschoot7 has found that EU interest rates are closely related to US interest rates. More importantly he a showed that while EU and US firms are not exactly similar by nature ,their 5-7 years A rated bond spreads follow a similar trend ,prior to 2003. This finding allows us the employ the diff in diff regression on 2 different groups without the threat of validity hazard, due to the nature of this methodology .For the control group I will use a list of European based corporation traded in Frankfurt, Brussels, Lisbon, Amsterdam, London, Paris, and Swiss stock exchanges, which are not cross listed and traded in either NYSE or NASDAQ or. That is because cross listed firms must comply with SEC regulation and adopt SOX as well, and implementing these firms will result in an internal validity hazard to my estimation- inability to correctly estimate the effect of SOX had on the price of debt because it cannot distinguish between observations that had gone through the treatment to those who did not
7
23 The regression equation is then as follows:
1) With being a Dummy variable equal to 1 if observation is in the time period after SOX was enacted, and therefore comply with SOX regulation., and 0 otherwise, is a Dummy variable equal 1 if it is U.S. firm and 0 otherwise. Is the interaction term between and and is the diff in diff estimator. Is a list of financial controls. The dependent variable in this estimation is ,which is the difference between government issued bonds and firms issued bonds, with the same maturity date. If my hypothesis is correct, I expect
to have a negative significant sign, proving that SOX had reduced the spread of US. Firms bonds, compared to European firms bond spreads. This in turn will prove as an indication that better corporate governance lowers firm's cost of debt.
The list of firm financial control variables is as follows:
Firm leverage, or which is defined as: . We know from previous empirical and theoretical research that the debt agency problem is more pronounced in highly leveraged firms. It is therefore plausible that the introduction of SOX had a stronger effect on highly leveraged firms. To control for this we have include this variable.
Firm profitability controls : earning per share which is defined as:
, and return on equity which is defined as: . The reason I include these 2 indicators is that while earning per share is a reliable indicator for firm profitability, it's a function of firm's number of shares, which is different between firms. Return on equity is not dependent on number the number of shares, and can be used to
24
Market value, which is defined as price*number of outstanding shares. This is
to scale the observations. Bigger companies with higher market value might have better access to credit, thus lowering their cost of debt. Adding the market value variable controls for this difference.
Industry - an indicator variable that represents the firm's industry in which it is active and accepts: 1 for industrial, 2 for utility, 3 for transportation, 4 for bank/saving loans, 5 for insurance and 6 for other financial institution.
To control for differences in bond characteristics between samples, I will include this controls: Duration of debt, which is Macaulay duration and Debt age, which is the period since the bond was issued. This will also neutralize any changes in the trend in both the control and test group.
The same methodology will be used to examine the effect of SOX on firm's debt to capital ratio. This will be done by employing a similar diff and diff, only this time then the dependent variable is , which is debt to capital:
2)
Presumably by lowering the price of debt, SOX has a positive effect on the debt to capital ratio in US firms. If that is true, I expect to be significantly positive.
Data on both bond yields and financial controls is taken from Thomson Reuters DataStream database.
25
4. Data Sample
To construct the difference in difference estimation, we had to chose the cut-off point between the 2 time periods- the before and after the natural experiment. As explained in the methodology section, the SOX act constitutes a solid natural
experiment. It was passed in July 15 2002. However firms fully adopted it only a year later, by 2003. Therefore we choose the first time period to be the year 2002, or to be precise the date 1/1/2002. Note that this is before SOX was enacted (15th of July), in order to capture any market anticipation that may had an effect on bond yields. Not doing so may result in estimation bias. The second time period is 1/1/2004, well after SOX was adopted.
Most of SOX provisions were effective immediately or over the course of 2003. However, companies were given more time to comply with section 404 (b) - companies assessment of its own internal controls. The extension was until 15th of July 2005. Knowing this, we chose to implement another dif and diff regression, in order to capture the effect the compulsory internal control report had on the price of debt. To do this we chose the first time period to be the same as before 1/1/2002. The second time period we chose to be 1/1/2006. Choosing a bigger time frame in the estimation ,however, can result in estimation bias due to noise from unrelated factors. This is why we chose to estimate the two regressions , and check for consistency between the two.
[INSERT TABLE 2 HERE]
For the sample of US Firms, I downloaded a list of financial data for all US traded firms in the NYSE and in the NASDAQ, in the dates 1/1/2002, 1/1/2004 and 1/1/2006, from Thomson Reuters DataStream database. This was done to insure as large as possible sample. I then matched this data with data on firm's bond spreads issued by the same companies, at the same dates, also from DataStream database. In a case when there are multiple bonds issued by the same company, I matched each one of them with its firm financial controls. Convertible bonds were not included in the sample. This is
26 because they are mechanism of minimizing the debt agency problem themselves. By not including them in our sample we can control for their effect on bond spreads. After clearing for missing data, I am left with 132, 132, and 131 observations for each time period respectively (also see table 2). There is a notable decrease on mean spread between year 2002 and subsequent years: from 247 basis points in 2002 to 194 basis points in 2007. The smallest company in market value terms in the US sample is 28.02 million dollars in the year 2002, and the largest is 129,839 million dollars, also in 2002. A firm with a market value of 129,839 can be considered a large firm, by US market
standards. Also there is a big variance in company size in the sample – the standard deviation in firm size in each year is approximately 23,000 million dollars.
[INSERT TABLE 3 HERE]
For the European sample, I downloaded all the financial data for companies traded in these following European stock exchanges: Frankfurt London Paris Amsterdam Brussels Swiss , in the dates 1/1/2002, 1/1/2004 and 1/1/2006, from Thomson Reuters DataStream database. Again I did this in order to insure as big as possible data sample, also due to the fact that there is less data available on European markets. Next I
matched bond data – spread life and duration, with the firm financial controls, for each time period.
Foreign companies, which are traded in NYSE and NASDAQ, are subject to SEC ruling and regulations ruling, and as such must comply with the SOX act. In order to make sure none of my observation in the European sample is a cross-listed firms, I cross checked my sample with a list of foreign traded companies in the US, published by the SEC for each year8. After clearing for missing data, I am left with 54,54 and 42
observation of non cross listed firms for the years 2002, 2004 and 2006 respectively.(see also table 3).
8
27 Similarly to the US sample, there is a decrease in European mean bond spread between the years 2002- 2007: from 298.81 basis points to 158.62 basis points. Duration is in the same magnitude in both the US and Europe sample. Market value of the average European firms in the sample tends to be smaller than its US counterparts - in the range of 4216 to 7483 million Euros, compared with a range of 13034-15848 million dollars in the US sample. Also biggest company in the sample, for each year is 28550, 27857, and 38039 million Euros which is, is about one quarter of the size of the biggest US firm. Debt to capital ratio seems to be bigger in European firms then US firms: about 50% compared to 40 in average respectively, meaning, that European firms are more levered then their US counterparts in our sample.
[INSERT TABLE 5, 6, 7 ,HERE]
Next I checked to what degree the estimated variable are correlated to each other, using correlations matrix table, for each regression. Duration and life of the bond are somewhat negatively correlated with bond spread, within each regression. However none of the independent variables is strongly correlated to another, meaning that there is no multicollinearity9problem in my estimation, which may affect its validity.
Finally, most of the firms in my sample, in all of the years, are industrial firms. They constitute about 60% of the sample. Financial institutions are then next big group, amounting to 13% of the sample.
9 Multicollinearity is a phenomenon when one or more independent variable in the estimation is a linear
combination of the other variables, or can be estimated by such a linear combination. This affects the precision of the estimators, and increases their standard deviation.
28
5. Empirical Finding and Results
[INSERT TABLE 8 HERE]
The first estimation, regression number 1, is the estimating the effect SOX had on cost of debt, with the cutoff date being 1/1/2004. Everything after 1/1/2004 is considered to have undergone the treatment, and is affected by SOX. Everything before 1/1/2002 did not go through the treatment. The diff in diff estimator is the interaction term US*SOX, and is significant at 1%, and highly negative: -224.6 basis points. That amounts for a decrease in bond spread of more than 2% caused by SOX. Also the adjusted is pretty high at 0.712. The regression is also highly significant, with F statistic of 407.8. After adding the control variables Life and Duration, the
increases to 0.787. The diff in diff estimator drops to -132.4 basis points. However it is still significant at 1%. Adding the other financial controls - market value, earnings per share, return on equity, debt to capital ratio, and industry does not change the
regression by much, however only market value, earnings per share and debt to
capital controls are significant. The diff and diff estimator in this estimation is -100.7 basis points, which amounts for a 1% drop in bond spread due to SOX, and is significant at 1%. Also the regression is significant with an F statistic of 137.8. This leads us to the conclusion that there is a significant effect of the SOX act on the price of debt.
We come up with somewhat different results in estimation (4) (5) and (6), regarding the effect of SOX on firm's debt capacity. Surprisingly enough SOX caused a significant decrease in Debt to capital ratio in all 4 estimation: with and without the controls. In estimation (4) it amounts for a decrease of 55.19% in debt to capital ratio, however this is clearly a biased estimation. Adding bond and financial controls, SOX affect on Debt to capital ratio drops down to a decrease of -11.35%, and is significant at 5%. SOX, it would, seem caused a reduction in firms debt to capital ratio in this time period.
29 [INSERT TABLE 9 HERE]
Next I estimated the effect of SOX effect on bond Spread and Debt to capital, with the cut-off point being 1/1/2006. This is after the extended deadline in which all companies had to implement rule 404(b) ordering them to coduct the internal control report. In this estimation we check the difference between 1/1/2002 bond spread and debt to capital ration with 1/1/2006 bonds spreads and debt to capital ratio, on both European and US firms. Indeed, choosing a smaller time frame for this estimation would have allowed us to narrow down more precisely the effect rule 404 (b) of the SOX had on spread and price of debt. However, although companies could have waited with the implementation of rule 404(b) until the deadline, they may have chosen to implement it at an earlier date. Choosing a shorter time period would not have captured the full effect of this change, and may have resulted in what is called "attrition" in the treatment group, that is some US companies in the time period before SOX will have undergone the treatment, which will cause the estimation to not distinguish properly between companies who receive the treatment to those who did not. This will cause the diff and diff estimator to be biased. To prevent this from happening we chose the bigger time frame.
In estimation (1) we can see that the diff and dif estimator is significant at 1%, and has a negative effect on spread of 225 basis points. When adding the duration and market value controls this negative effect drops to 112.6 basis points. It is, however, significant at 1%. Further adding all of the financial controls the effect drops to -80.85 basis points, and is still significant at 1%. This amount a decrease of 0.8% in bond spread. However the SOX dummy variable is not significant anymore. This mean that as a whole, SOX had no effect on both European and US, firm aggregated together. It did, however effect US firms, evident by the diff and diff estimator being significant. All the estimations are significant as a whole, with an F statistic of 121.5 and above, and an of 0.685, 0.716 and 0.74, respectively. Similar to the estimation with which the cut- off point was
31 debt. However it is likely due more to the entire provisions of the act, and not due to rule 404(b) regarding internal control report itself, as it did not change much the magnitude of the negative effect in basis points in bond spreads.
Looking at estimation (4), SOX causes a decrease in of 59.68% in debt to capital ratio. This is clearly a biased estimate. Adding the controls duration and life in estimation (5), we get that SOX causes a decrease of 19.42% in the debt to capital ratio, and is
significant at 1%. However after adding the financial controls in estimation (6), the diff in diff estimator drops again to -10.78%, it is not statistically significant anymore. Also the of the regressions improves from 0.581 to 0.722. In light of these results it is questionable whether SOX had any influence on the debt to capital ratio in US firms, as the results are not statistically significant enough .Also they are not consistent in the 2 time periods tested. This may be a result of a limited data sample. Thus, in contrast to this paper hypothesis, we argue that corporate governance has no effect firm's debt to capital ratio.
It is worth noting that in all time periods, being a US firm increases the bond spread and debt to capital ratio, evident by the indicator variable US being positive and significant at 1% al almost all regression. Keep in mind that in our sample European firm were more leaved, which rules out bankruptcy costs as explanation. This however is true to our small sample of European firms, and a much large sample is needed to be able to say this with absolute certainty.
The results of this paper show that there is indeed a plausible effect of corporate governance on the price of debt. In many ways, this follows the same conclusion from previous papers. Anderson Mansi and Reeb (2004) found that certain board
characteristics, a major part of firm corporate governance system do indeed reduce bond spreads. Klock, Mansi and Maxwell(2005) again found that percentage of institutional investors in the firm reduces the bond spread. Bhojraj, Sanjeev and
Partha(2003) have come up with similar results. Meanwhile Amir Guan and Livne(2009) found that SOX itself increase auditor independence, while auditor independence itself
31 decreases the cost of debt, implying the SOX act had decreased the cost of debt, similar to our results. The implications for this findings are vast, and will be discussed in the closing chapter of this paper.
32
6. Consistency checks.
In previous section we have showed empirically that SOX act had statistically significant negative effect on bond spreads in US firms. This suggests that improvements in corporate governance can indeed reduce the firm's cost of debt. However this paper finding validity may suffer from 2 reasons: first my European firm sample is rather small, and second US and EU firms are not entirely similar, and therefore the diff and diff regression itself might be biased. To solve this we approach it from 2 different angels- we test my initial hypothesis on US firms alone, using an unbalanced panel data
regression, and we test my hypothesis on EU firms alone, using a methodology similar to Arping and Sautner(2010)
My first test in the unbalanced panel data. SOX here is an indicator variable, accepting 0 if observation is before 15th of July 2002, and 1 otherwise. The 380 observations are taken from the dates 1/1/2000, 1/1/2001, 1/1/2002 and 1/1/2004, 1/1/2006 and 1/1/2007. While this sample is a similar in size to the number of observations in our diff in diff estimations, in a panel data methodology consisting of only US firms it is sufficiently large to be robust. The estimation results are brought table 10.
[INSERT TABLE 11 HERE]
Estimating the effect of SOX act in bond spreads using only US firms sample it appears that similarly to our previous estimations, SOX does have a statistical significant reduction of bond spread of about -55.55 basis points, which amount to a decrease of 0.55% (estimation 1), and is significant at 1%. Adding the financial controls the effect drops to -38.63 basis points, or -0.37% reduction in bond spread, and is still statistically significant. These results are consistent with the previous diff and diff regression, albeit at a lower magnitude, meaning that they are robust.
In the estimation of the effect of SOX on debt to capital ratio, we arrive at a similar conclusion as our diff and diff estimations. In estimation (4) SOX is both negative
33 (-6.202) and statistically significant at 1%. However the regression F statistic is quite low,3.1, and so is the adjusted . This means that the regression, as a whole, is not statistically significant. Therefore according to our panel data estimation SOX had no effect on the debt to capital ratio in US firm. This is similar to our results from diff and diff estimations, and proves that they are robust. However this neither validates my second hypothesis, that improved corporate governance increases firms leverage ratio, nor does it confirms Jensen(1986) theory of debt as a disciplinary method.
[INSERT TABLE 10 HERE]
The second robustness check we employ is checking whether the same decrease in bond prices happened in European cross listed firms compared to non cross listed firms. This is because European firm are similar in nature, as opposed to US and European firms, and this may improve our estimation validity. Using methodology similar to Sautner and Arping(2010) we compare European cross listed firm (in NYSE and NASDAQ) , who had to adopted SOX in accordance with SEC listing rules, to non cross listed European firms, which did not, with the cutoff point being 1/1/2004. The results are brought in table 10. As stated in the methodology section, obtaining data about bond spreads in EU firms was difficult, and as a result the sample consist of 86 firms. To add robustness we have taken monthly observations in the year 2002 and 2004. After cleaning for missing data this resulted in 2644 observations. Similarly to our previous estimation, there is a statistically significant decrease in cross-listed firms spreads due to SOX, of about 60 basis points. Also there seems to be a decrease in leverage ratio, of about 16% caused by SOX, which is significant. This shows that our finding are a good approximation to the effect SOX had on the price of debt.
All in all, employing different estimation methodology of fixed effects panel data regression, and diff in diff regression using EU firms we arrive at the same results and conclusion as our previous estimations. This means that this paper results and
34
7. Conclusion and Final remarks
In this paper we sought to empirically asses if and how much can good corporate governance decrease the cost of debt. Existing literature, and corporate finance models, suggests that there should be a connection between the two. The debt agency costs are caused by risk shifting and underinvestment problems, and both are linked to an
increase in the firm's price of debt. In these cases equity holders expropriate the
creditors, either by investing in risky assets, and thereby transferring the risk to the debt holders, or by not investing in safe positive net present value assets at all, thus
decreasing creditors wealth. Current models suggest that rational investors are well aware of these phenomena, and therefore are willing to lend to firms with such problems at a discount. To test empirically whether these assumptions are true in the real world, we construct differences in differences model, comparing US firms who had adopted the Sarbanes and Oxley act to and European companies who did not. The SOX act constitutes a major change in the quality of corporate governance, following the Enron scandal. It improves, among other thing, the quality of financial disclosures and company internal controls. Using this I tried to estimate exactly the difference, in both bond spread and debt to capital ratio, that this change in the quality of corporate governance had. Improving on prior literature, my model is more robust to omitted variable bias and endogeneity problems.
Our initial hypothesis was that indeed US firms after SOX have experienced a decrease in bond spread, and an increase in debt capacity. Our empirical results show that indeed the SOX act reduced bond spreads in US companies by a considerable amount, of an average around 1%. This effect happened even before the adoption of rule 404(b), necessitating a company to conduct a report of its own internal control, in itself an important corporate governance measure. In many ways this results is similar to previous papers findings: Anderson Mansi and Reeb (2004) Klock, Mansi and Maxwell (2005) and Bhojraj, Sanjeev and Partha Sengupta (2003) have found that board
35 structure and size effect bond spread. Andrade, Sandro, Gennaro, and Hood (2009), have found a more direct link between SOX and bond CDS prices, while Amir Guan and Livne (2009) have shown that SOX increase auditor independence, and this in turn decreases bond spreads.
In contrast we find no evidence that SOX increased the debt to capital ratio. Both our panel data and diff in diff estimation were statistically insignificant, meaning that we cannot confirm our second hypothesis, nor can we prove Jensen (1986) theory of debt as being a control mechanism on the management by the shareholders. This may have to do with our limited data sample, and should be tested further.
This paper results help to shed light on the importance of good corporate governance on the price of debt, and consequently on firm capital structure as a whole. In light of this paper findings, it is imperative that firms adopt measures to increase the quality of their corporate governance. This will reduce inefficiencies, help them to reduce their financing expenses, increase their growth potential and benefit the economy as a whole.
However our results must be taken with caution. First, the empirical data was taken from the years 2002-2007. These were the years before the current financial crisis, when cheap credit was available due to low interest rates offered to commercial banks by the central banks. This might have a considerable effect on these paper findings. In the years of the economic boom, it is reasonable to assume that bond
spread will decrease as well, regardless of SOX. Secondly, due to missing data, this paper sample was very small, and especially the European sample was very small, due to the difficulties finding data on European markets. This may hurt the validity of the
difference in difference estimation, and a bigger sample is need for a more clear answer. And third, due to the limitation of my methodology, there might be a different factor causing the decrease in bond spreads and debt to capital ratio, which was happening in those same years, and was captured by my estimation. In other words, there might be
36 another "natural experiment" occurring on the same time, acting on both the treatment and non treatment groups, which was captured by the diff and diff estimator.
In light of these problems, it would be interesting to see, in a further research, the effect corporate governance has on the price of debt , with a data sample in the years after the financial crisis- that is, in the years 2008 and beyond. This may show the difference the financial crisis, with all changes it brought, had on both the cost of debt, and the quality of corporate governance. It will also be beneficial to conduct similar research, using similar methodology, comparing different corporate governance systems: the German 2 tier board system, the Dutch system, Chinese state owned enterprise, etc, and to see how this differences impact the cost of debt and capital.
37
8. Bibliography
Aharony, Joseph and Itzhak Swary. "Quarterly Dividend and Earnings Announcements and Stockholders' Returns: An Empirical Analysis." The Journal of Finance 35, no. 1 (Mar., 1980): 1-12.
. "Auditor Independence and the Cost of Capital before and After Sarbanes–Oxley: The Case of Newly Issued Public Debt." European Accounting Review 19, no. 4 (2010): 633-664. Anderson, Ronald C., Sattar A. Mansi, and David M. Reeb. "Board Characteristics, Accounting
Report Integrity, and the Cost of Debt." Journal of Accounting and Economics 37, no. 3 (2004): 315-342.
Andrade, Sandro, Gennaro Bernile, and Frederick Hood. "SOX, Corporate Transparency, and the Cost of Debt." Available at SSRN 1257002 (2009).
Arping, Stefan and Zacharias Sautner. Did the Sarbanes-Oxley Act of 2003 make Firms Less Opaque?: Evidence from Analyst Earnings Forecasts Tinbergen Instituut, 2010.
ASHBAUGH-SKAIFE, HOLLIS, DANIEL W. COLLINS, WILLIAM R. KINNEY JR, and RYAN LAFOND. "The Effect of SOX Internal Control Deficiencies on Firm Risk and Cost of Equity." Journal of Accounting Research 47, no. 1 (2009): 1-43.
Bhojraj, Sanjeev and Partha Sengupta. "Effect of Corporate Governance on Bond Ratings and Yields: The Role of Institutional Investors and Outside Directors*." The Journal of Business 76, no. 3 (2003): 455-475.
Bradley, M., G. Dallas, E. Snyderwine, and D. Chen. "The Effects of Corporate Governance Attributes on Credit Ratings and Bond Yields." Available at SSRN 1327070 (2008). Brennan, Michael J. and Eduardo S. Schwartz. "The Case for Convertibles*." Journal of Applied
Corporate Finance 1, no. 2 (1988): 55-64.
Chen, Kevin CW, Zhihong Chen, and KC John Wei. "Agency Costs of Free Cash Flow and the Effect of Shareholder Rights on the Implied Cost of Equity Capital." Journal of Financial and Quantitative Analysis 46, no. 1 (2011): 171.
Cremers, K. J. Martijn and Vinay B. Nair. "Governance Mechanisms and Equity Prices." The Journal of Finance 60, no. 6 (Dec., 2005): 2859-2894.
EISDORFER, ASSAF. "Empirical Evidence of Risk Shifting in Financially Distressed Firms." The Journal of Finance 63, no. 2 (2008): 609-637.
Fama, Eugene F. and Merton H. Miller. The Theory of Finance. Vol. 3 Dryden Press Hinsdale, IL, 1972.
GIROUD, XAVIER and HOLGER M. MUELLER. "Corporate Governance, Product Market Competition, and Equity Prices." The Journal of Finance 66, no. 2 (2011): 563-600.
