T HE INFLUENCE OF BOND ISSUES ON THE SPREAD OF A CREDIT DEFAULT SWAP

T

HE INFLUENC E OF BOND ISSU ES ON THE SP READ OF A CREDIT

DEFAU LT SWAP

University of Groningen

Master of Science in Business Administration Specialization: Finance

Liselot Lubberman Student number: 1536060

Supervisor: prof. dr. L.J.R. Scholtens 2nd Supervisor: dr. J.O. Mierau

L.T.M. Lubberman 3

Abstract

This study investigates the determinants of the spread of a credit default swap (CDS). According to the structural model, default is influenced by volatility, risk free rate and leverage. However, these variables are not able to explain the complete spread of a CDS. This research adds the event of a bond issue as a new determinant. This research is structured using panel data consisting of 187 firms and covering 60 months. The influence of a bond issue is investigated in combination with the maturity of the issue. The business cycle during the issue and the sector in which the issue took place are used as circumstances that possibly influence the CDS spread. A bond issue during a cyclical downturn and an issue in the energy sector turned out to be significant. However, the coefficients of a bond issue are small as well as the added explanatory power.

Key words: Credit default swaps, determinants, credit spread, bond issue.

L.T.M. Lubberman 4

1.

I

NTRODUC TION

Credit default swaps (CDS) are a relatively young derivative. As suggested by the International Swap and Derivative Association, it is only traded in notable amounts since 2001. In the years following, its popularity faced a tremendous growth, with increases in the traded amount up to 128% per year in 2004 (www.isda.org). Since mid 2008 this trend is more or less reversed and the market is slightly declining. The credit default swap is the most commonly traded credit derivative. A credit derivative is, as described by Ericsson et al. (2004), a contingent claim that allows the trading of default risk separately from other sources of uncertainty. The most important participants in the credit derivative market are banks, hedge funds, corporations, pension funds and insurance companies. The great advantage of making use of credit derivatives, as stated by Greatrex (2008), is that it reduces the credit exposure while maintaining the underlying asset and avoiding certain legal, regulatory and tax issues.

The essence of a credit default swap is a derivative that provides insurance against default of the particular company or sovereign entity. This party is also known as the underlying entity or reference entity. The buyer of a CDS will make fixed and regular payments to the seller and in return obtains the right to sell the bond issued by the underlying entity at its face value if a credit event occurs. A credit event is usually defined as a change in the counterparty’s ability to perform its obligations. This change might be due to the failure to make a payment on a debt obligation, a restructuring or bankruptcy.

L.T.M. Lubberman 5 price fluctuations are discovered earlier in the CDS markets. It thus turned out that CDS spreads are a superior measure of risk (Blanco et al., 2003).

In this study I want to find out what factors determine this CDS spread and whether the explained part of the CDS spread can be improved by adding one additional variable. The main variable investigated in this research is the event of a bond issue but also the influence of the maturity of a CDS will be investigated. A bond issue may influence the risk on a corporate bond, it may also has an effect on the credit spread of a CDS. The maturity of a bond influences the bond credit spread, consistently it might have an impact on CDS spreads. The articles of among others Collin-Dufresne et al.(2001), Avramov et al. (2007), and Ericsson et al. (2004) try to explain credit spreads by using the so called structural model. The structural model, as suggested by Merton (1974), assumes that a firm’s default is directly influenced by its economic fundamentals. These fundamentals are usually the amount of leverage of the underlying entity, equity volatility of the underlying entity and the risk free rate of the country the underlying entity is located. However, researches based on these variables are only able to explain about 25% to 30% of the observed credit spread (Collin-Dufresne et al. (2001), Avramov et al. (2007), Ericsson et al. (2004)).

By adding the event of a bond issue as a variable, this research tries to find out whether a bond issue influences the CDS spread in a significant way. When looking at the event of a bond issue this can have different implications. An increase in the leverage position of a firm might add substantial risk of default however, it can also lead to tax gains that make it rewarding to issue debt. What effect will occur depends on the circumstances a firm finds itself in during a debt issue. Circumstances that in this research are identified as possibly having an influence on the credit spread of a CDS are the business cycle a firm finds itself when issuing the bond and the capital intensity of the sector the firm operates in. With respect to the influence of cyclical differences under which a bond is issued, the hypothesis is based on theory provided by, among others, Moore (1956). Moore (1956) states that corporate bonds face during a time of cyclical downturn a larger credit spread than bonds during cyclical upturn. This research hypothesizes that this effect also holds for CDS spreads during a bond issue. With respect to the capital intensity of the sector the firm operates in on a bond issue, theory of among others Belkaoui (1980) is used. Belkaoui (1980) states that the more capital intense the sector is, the larger credit spread is on its corporate bonds. This research hypothesizes that this effects also holds on CDS spreads during a bond issue.

L.T.M. Lubberman 6 sample as well as the influence of the maturities on the scenarios are tested. With respect to the maturity of a bond issue the hypothesis is based on an article of Elton et al. (2001) that state that bonds issued with a short maturity face a larger spread on its corporate bond spread than bonds issued with a longer maturity. In this light, bonds issued with a short maturity, between 3 and 7 years, are expected to face a larger credit spread on its CDS than a bond issued with a maturity of 10 years.

This research follows the fundamentals provided by the research of Ericsson et al. (2004). Ericsson et al. (2004) investigated the influence of the variables implied by the structural model on the credit spread of a CDS. They found that the variables leverage of the underlying entity, equity volatility of the underlying entity and risk free rate, measured as the swap rate of the country the underlying entity is located, have a significant effect. Based on these results this research uses these variables as control variables and adds the event of a bond issue as a new variable. To identify the influence of these variables a panel data regression is conducted. The panel data set consists of 187 companies and stretches a time period of 60 months.

After running the proposed regressions it turned out that an issue during a low business cycle does have a significant effect on the CDS spread but shows another sign than expected. A bond issued during cyclical downturn faces a lower credit spread. One possible explanation for this phenomenon is that a bond issued during economic downturn may signal positively to investors that the firm did not had to issue equity, which is -according to the pecking order theory of Donaldson (1961)- the least optimal form of financing. An issue in the energy sector also has a significant influence on the CDS spread. As being the least capital intense sector a bond issue in the energy sector suppresses the change in CDS spread most relative to the other sectors, which is conform theory. Also the explanatory power of the model shows a small improvement The regressions that do find significant results, have very small coefficients and add only a little increase of explanatory power of the model. Therefore it can be stated that a bond issue appears to be not very important in determining the variables that influence the changes in CDS spread. Dividing the data between credit default swaps with a short maturity and a long maturity did not increase the fit of the model in terms of significance of the coefficients or in remarkable improvements in the explanatory power of the model.

L.T.M. Lubberman 7 The remainder of this paper is structured as follows: section 2 provides a short overview of the existing literature. Also the investigated hypotheses are explained in this section. Section 3 describes the methodology this research uses and describes the tested hypotheses. Section 4 gives an overview of the collected data, describes the used variables and the descriptive statistics. The results of this research are presented in section 5. The last section provides concluding remarks together with a discussion.

2.

L

ITERA TURE

R

EVIEW

Recently a large amount of literature has been developed on the pricing of so called credit sensitive instruments. The most important type of credit sensitive instruments is the corporate bond although credit derivatives have gained importance lately. This section first discusses the existing theoretical literature relevant to this subject. It is followed by the empirical literature and ends with a close description of the tested hypotheses in this research.

2.1

T

HEORETICAL LITERATURE

There are several frameworks for valuing risky assets. The most common division is made between the so called reduced form models and the structural models. The reduced form model is suggested by among others Duffie and Singleton (1999). As described by Chen et al. (2008), these models assume that defaults occur unexpectedly and follow a Poisson distribution. Abid and Naifar (2006) add that reduced form models estimate the risk neutral probability of default over a given interval with actual credit spreads, without the necessity of knowing the underlying cause of default. The dynamics of default are thus exogenously determined and market data is necessary to recover the parameters needed to value credit sensitive claims (Ericsson et al., 2004). This approach allows for the pricing of a credit sensitive instrument. However, it does not allow for a closed form solution since this model cannot estimate all the parameters in the model simultaneously because of computational difficulties (Chen et al., 2008). Besides, these models also remain silent on what theoretical determinants to include (Ericsson et al., 2004). These implications of the reduced form model make it hard to find the practical use of this model.

L.T.M. Lubberman 8 essence of a structural model, as Abid and Naifar (2006) state, is that default is defined as a contingent claim by describing the reasons of default and then price the security following the option pricing technique of Black and Scholes (1973). The firm defaults when the value falls below this determined default threshold. The most common implied variables by the model of Black and Scholes (1973) are the risk free rate, leverage and volatility as they have a clear impact on default probabilities. This impact with of volatility and leverage on default risk is positive meaning that the higher the volatility or the leverage, the greater the risk of default. This impact of risk free is negative meaning that the higher the risk free rate, the lower the probability of default. However, as stated by Ericsson et al. (2004), the structural model have been plagued by a rather poor performance in empirical studies when using only the variables risk free rate, leverage and volatility. Avramov et al. (2007) state that the success of the structural model is closely connected to choosing the correct variables that capture the variation in credit spread changes. For this reason a number of empirical researches has emerged to find out what variables besides risk free rate, volatility and leverage influences the credit spread. These researches are discussed more thoroughly in the next paragraph.

2.2

E

MPIRICAL LITERATURE

L.T.M. Lubberman 9 Investment grade bonds are bonds with a credit rating of BBB or higher. Avramov et al. (2007) state that company level fundamentals and common factors are important determinants of credit risk changes among high yield bonds, but to a lesser extent among investment grade bonds. Another reason Avramov et al. (2007) give for the differences is that they use slightly more variables to mention idiosyncratic equity volatility and stock return momentum. Stock return momentum is defined as the predictability of equity returns from past returns. Higher momentum in equity returns implies a higher future valuation and a lower probability of default (Avramov et al., 2007). To include equity volatility is based on research of Campbell and Taksler (2003). Campbell and Taksler (2003) state that equity volatility can explain as much cross-sectional variation in yields as can credit ratings. Furthermore, equity volatility also helps to explain the longer term trend of the corporate yield (Campbell and Taksler, 2003).

Recently a substantial amount of empirical work focusing solely on credit default swap spreads has emerged. Fundamental research is provided by Blanco et al. (2003). Blanco et al. (2003) state that when the swap rate is used as the risk free rate, the CDS spread lies quite close to the bond yield spread. They also find that the CDS spread incorporates new information faster than bond spreads. Longstaff et al. (2005) follow a different approach by using the information in the CDS spread to obtain a measure of the size of the default component of the corporate bond spread. Longstaff et al. (2005) find that the majority of this spread is determined by default risk. Berndt et al. (2004) takes these findings further and uses CDS spreads to estimate the height of the default risk premia of US corporate debt. Hull et al. (2004) discusses the impact of credit rating announcements on the spread of a credit default swap. The results of Hull et al. (2004) are more or less conform the results of Blanco et al. (2003) by stating that CDS spreads predict negative rating events. According to among others Longstaff et al. (2005) and Hull et al. (2004) the vast majority of the CDS spread is determined by factors affecting default. Therefore it becomes apparent to determine the factors affecting default. Ericsson et al. (2004) focused on a minimum set of determinants to mention the swap rate as risk-free rate, equity volatility and firm leverage and tested whether changes have a significant influence on the credit spread. All these variables turned out to have a significant effect on the credit default swap spread. The explanatory power, measured as the R-squared, of the variables risk free rate, volatility and leverage on the research of Ericsson et al. (2004) was 23% and thus more or less conform the results of Collin-Dufresne et al. (2001) on corporate bond spreads. Greatrex (2008) does a comparable research adding a few additional variables like firm value and credit rating resulting in a slightly higher explanatory power of 30%.

L.T.M. Lubberman 10 regime dependant and sector specific behavior. With regime dependant Alexander and Kaeck (2008) mean that the dependence between credit spread and their determinants depends very much on the market circumstances with equity volatility being of greater influence in volatile markets than in tranquil markets. Also, the explanatory power in volatile markets was significantly higher than in tranquil markets. With respect to sector specific behavior a distinction is made between financial firms and non financial firms. It is stated that CDS spreads on financial firms are immune to interest rate changes but that there is a negative association between CDS spreads and interest rate changes on non financial firms. My research proceeds on the research of Alexander and Kaeck (2008) although my division in sectors is based on another scale. My research only covers sectors in the non financial industry to mention non-cyclical consumer goods, transport and energy. Since these sectors are all located in the non financial industry a negative relation between interest rates and CDS spreads is expected.

Ericsson et al. (2004) concludes by stating that a large fraction of the variation in spread differences remains unexplained. This research aims to decrease this unexplained part by adding a new variable to mention a bond issue and the maturity of the CDS. By finding out whether these variables have an influence on the credit default swap spread, the explanatory power might increase, leading to a more suitable model.

2.3

H

YPOTHESES

L.T.M. Lubberman 11 However this theorem only holds in a stylized world without for example taxes, bankruptcy cost and agency costs. In the real world these costs do exist. As described by, among others, Collin-Dufresne (2001) and Greatrex (2008) when the leverage position of a firm increases sharply this leads to an increase in bankruptcy costs and an increased probability of default. This increased probability of default also leads to an increased credit spread. As stated by Campbell and Taksler (2003) volatility of a firm hurts bondholders, because it increases the probability of default. They also state that equity volatility explains as much variation in corporate credit spreads than do credit ratings. Thus the higher the equity volatility the higher the credit spread on corporate bonds. Ericsson et al. (2004) provided evidence that this effect also counts for credit spreads on credit default swaps.

In this research the number of determinants is extended with the event of a corporate bond issue. There might be several different reasons for a firm to increase the debt position by means of a bond issue. Following the so called static capital structure theory (Modigliani and Miller, 1961) firms optimize their capital structure period by period. This theory states that there is an optimal structure between having a large amount of debt when business is underperforming and bankruptcy costs and default risk become high, and having a substantial amount of debt when business is going well which results in the tax benefits of debt and lower default risk. Another theory worth considering is the pecking order theory (Donaldson, 1961) which assumes that firms have a general funding preference. This theory states that firms prefer funding with retained earnings over external financing. If these funds are not sufficient they will first try to issue debt. An equity issue suits as a last resort. According to the pecking order theory a debt issue during poor times might reduce the risk, since this debt issue makes it clear that the firm does not need to issue equity yet. This can then be seen as a positive sign, and thus risk reducing. The opposite happens when a firm asks for debt in prosperous times. An issue then gives the sign that the firm is not able to finance itself through retained earnings. This can be seen as a bad sign which increases the risk. Since the position of debt in the pecking order theory is not in a definite position and may lead to different signs and because of the fact that there is no optimal debt structure in the static capital theory it is likely that a bond issue does not always result in the same outcome. The effect of a bond issue will thus be dependant on the circumstances under which the bond is issued. To determine the effect of a bond issue on the CDS spread, different circumstances under which a bond is issued are identified. The circumstances under which an issue takes place or scenarios discussed in this research are based on cyclical differences and sectoral differences.

L.T.M. Lubberman 12 respect to maturity differences the article of Elton et al. (2001) states that there is a general tendency for the credit spread on corporate bonds to increase as the maturities lengthens. In other words, bonds that are issued at a short maturity are expected to face a lower credit spread and thus less risk, than bonds issued at a long maturity. It is expected that this result will also hold for credit spreads on credit default swaps. This research compares bonds issued with a maturity between three and seven years, with bonds issued with a maturity of ten years. The hypothesis is then as follows:

H1: A CDS with a short maturity faces a significant lower credit spread on credit default swaps, than a CDS with a longer maturity.

With respect to cyclical differences it is expected as stated by Guha and Hiris (2002) and also earlier by Moore (1956) that the credit spread behaves counter-cyclically. In other words, the spread becomes larger during economic recessions and smaller during economic expansions. Thus, when the event of a bond issue occurs, a division will be made between bond issues that take place during a recession and bond issues that take place during economic expansion. A recession is defined according to the classical definition of Shiskin (1974) as two following quarters of economic downturn. An economic expansion is defined as two following periods of economic upturn. This first group is expected to face an increasing credit spread, the second group is expected to face a decreasing credit spread. The hypothesis will be as follows:

H2: A bond issue during a recession is expected to increase the observed credit spread of the credit default swap.

H3: A bond issue during an economic expansion is expected to decrease the observed credit spread of the credit default swap.

L.T.M. Lubberman 13 expected that a bond issue in a sector with a higher capital intensity will lead to a higher CDS spread. The hypothesis examined is thus:

H4: A bond issue in the energy sector will have a lower credit default swap spread than the other sectors.

H5: A bond issue in the transport sector will have a higher credit default swap spread than the other sectors.

Variable Motivation

Risk free rate

-The higher the riskfree rate, the lower the default risk and thus CDS spread

Ericsson et al. (2004), Collin-Dufresne et al. (2001)

Leverage +

The higher the leverage, the higher the default risk and thus the CDS spread

Ericsson et al. (2004), Collin-Dufresne et al. (2001)

Volatility +

The higher the volatility, the higher the default risk and thus the CDS spread

Ericsson et al. (2004), Collin-Dufresne et al. (2001)

CDS short maturity

--CDS long maturity

-Bond issue Business cycle

High business cycle + Bond issue during the high business cycle increases the credit spread of the CDS

Low business cycle - Bond issue during the low business cycle decreases the credit spread of the CDS

Sector

Consumer

--Energy

-Transport

---Table 1 Literature; A description of the tested variables, their predicted signs, their motivation and previous literature in this topic.

Corresponding literature

Guha and Hiris (2002) and Moore (1965)

Belkaoui (1980) and Pinches and Mingo (1973) Predicted

sign

The shorter the maturity the lower the credit spread of the CDS

The more capital intensive the sector of the bond issue, the higher the expected credit spread of the CDS.

Elton et al. (2001)

3.

M

ETHODOLOGY

L.T.M. Lubberman 14

3.1

P

ANEL ANALYSIS

Conform research of Campbell and Taksler (2003) as well as Ericsson et al. (2004) to some extent this research is structured as a panel data analysis. Panel data has the differentiating aspect of containing a cross-sectional dimension as well as a time-series dimension. As suggested by Brooks (2008) one of the most important features of panel data analysis is that it becomes possible to tackle more complex problems by combining the time series dimension with the cross sectional dimension. With respect to this research this means that simultaneously the time series of 60 months are tested against the 187 firms as cross sections. Since there is an observation available for every unit and for every time period, the data can be classified as a balanced panel.

In panel data estimators, as stated by Brooks (2008), there are two approaches that are useful for financial research, to mention fixed effects models and random effects models. The essence of a fixed effect model is that the intercept in the regression model is allowed to differ cross sectionally but not over time, while all of the slope estimates are fixed both over time and cross sectionally which results in a constant slope (Brooks, 2008). With respect to this research this would mean that every firm could have another intercept due to firm specific situations. Another type of fixed effect models could have a constant slope with intercepts that differ over time but not cross-sectionally. This would mean that there are no firm specific differences but that there are changes part way of the sample period which influence all firms on more or less the same way. The general regression formula is reproduced in formula (1).The fixed effects model decomposes the disturbance term, uit of the regression equation into an individual specific effect, µi, and the remainder disturbance, υit, that varies over time and entity while capturing everything that is left unexplained about y (Brooks, 2008). This new formula is stated in formula (2).

Yit=α+βxit+uit (1)

Yit=α+βxit+µi+υit (2)

L.T.M. Lubberman 15

Yit=α+βxit+εi+υit (3)

Whether a fixed effect model or a random effects model is more appropriate is not always clear. As stated by Brooks (2008) the random effect model is more appropriate when the entities in the sample are selected randomly from the population, the fixed effect model suits better when the entities constitute the entire population. With the random effect model fewer parameters need to be estimated which saves on the degrees on freedom. The major drawback with the random effects model is that it only works properly if the error term is uncorrelated with the explanatory variables. To test whether the error term and the explanatory variables are uncorrelated a Hausman test can be conducted. As stated by Brooks (2008), if there exists correlation between the error term and the explanatory variables the parameters will be biased and inconsistent. If the null hypothesis of no correlation is rejected, a fixed effects model needs to be used. In this case, the fixed effects model on cross-sections showed the best fit and is also conform the research of Ericsson et al. (2004).

3.2

G

ENERAL MODEL

L.T.M. Lubberman 16 of this test can be found in appendix C. Studying differences instead of levels can also be explained on a theoretical base. As explained by Avramov (2007) a one-to-one relationship between bond prices and the credit spread exists whereas the credit spreads are directly associated by bond excess returns. Avramov (2007) concludes by stating that credit spread changes are the focus in bond pricing. The last assumption states that the errors follow a normal distribution. Finding out whether the errors are normally distributed a Jarque-Bera test can be done. If there is no normal distribution this can be corrected by taking the logatithmic values of the variables. In this research this last assumption is not adjusted for to remain comparable with the articles of, among others, Ericsson et al. (2004), Alexander and Kaeck (2007) and Collin-Dufresne et al. (2001). Since the amount of data is large, violating this assumption will not influence the results.

Another adjustment concerns using quadratic risk free rates. As stated by Collin-Dufresne et al. (2001) quadratic risk free rates can be used to capture non-linear effects due to convexity. This adjustment depends on the period stretched by the research, where the convexity will increase when the length of the investigated period increases. Following Collin-Dufresne et al. (2001) and also Ericsson et al. (2004), this research uses quadratic risk free rates.

3.3

T

ESTABLE RELATIONSHIPS

The general regression formula that is used in this research is described below and conform the formula used by Ericsson et al. (2004). According to theory CDS spreads are influenced by the risk free rate, measured by the swap rate of the country the firm is located, the equity volatility of the underlying entity and the amount of leverage the underlying entity incurred. In this research the influence of a bond issue on the spread of a CDS is also investigated as well as the influence of the maturity of the CDS on the CDS spread. The abbreviations used are ‘rf’ for risk free rate, ‘lev’ for leverage and ‘vol’ for volatility, ‘bondD’ for bond dummy, ‘lmD’ for long maturity dummy and ‘smD’ for short maturity dummy. Dummies are used in this research to determine whether bond issues influences CDS spreads as well as to determine the influence of the maturity on the CDS spread. Dummies are needed when the occurrence of an event needs to be measured since it is expected to shift the outcome. In this case, with respect to the first dummy

γ

bondD

L.T.M. Lubberman 17

(4)

(ten years) and a CDS issued with a short maturity (between three and seven years) which in that case receives a value of one and zero otherwise.

k c t i k c t i k c t i k c t i k c t i k c t i k c t i k c t i

smD

lmD

bondD

vol

lev

rf

CDSspread

ε

γ

γ

γ

β

β

β

α

+

+

+

+

∆

+

∆

+

∆

+

=

∆

C= high or low business cycle K = sector

I = cross-sectional firm T = time

In this research maturity is thus set as a dependant variable and constructed using a dummy variable. Dummy variables are used because the use of panel data with fixed effects conflicts with using maturity as a continuous variable. Also, to remain comparable to the existing literature (for example Collin-Dufresne et al. (2001)) I have chosen to model the maturities as a dummy to solve this conflict. For this research the maturities are divided in two subgroups to mention, credit default swaps with a maturity between 3 and 7 years (short maturity) and credit default swaps with a maturity of 10 years (long maturity). This division is comparable to the division used by Collin-Dufresne et al. (2001). Their short maturity sub-sample includes maturities between three and nine years, the long maturity sub-sample includes maturities for 12 years and longer. After running this regression it can be stated whether a bond issued with the specified maturity significantly influences the observed CDS spread.

One of the circumstances that might influence the CDS spread is the phase of the business cycle the bond is issued. In formula (4) this scenario can be found by the subscript C. The cyclical movement of firms is divided in two subgroups; the high business cycle, defined as two following periods of rising business cycle data and the low business cycle, defined as two following periods of falling business cycle data. The regression needs to point out whether a bond issue in the specified business cycle period influences the CDS spread. For this research two dummy variables are constructed. The first variable takes the value of 1 if the issue takes place in a period of cyclical upturn and zero otherwise. This dummy shows the effect of a bond issue during cyclical upturn. The second dummy takes the value of 1 if the issue takes place during cyclical downturn and zero otherwise. This dummy shows the effect of a bond issue during cyclical downturn.

L.T.M. Lubberman 18 intensity. Capital intensity is defined as the amount of debt related to the amount of invested capital. The sectors researched in this survey are; transport, energy and non-cyclical consumer goods. Energy is considered as being the least capital intense and transport as being the most capital intense. After running this regression it can be stated whether a bond issued in a specified sector significantly influences the CDS spread. With respect to this scenario again dummies are used to structure the research. To find out what the CDS spread is of a bond issued in the transport sector a dummy is constructed that takes a value of one if the firm that issues the bond operates in the transport sector and takes the value of zero otherwise. To find out what the CDS spread is of a bond issued in the energy sector a dummy is constructed that takes the value of one if the firm that issued the bond operates in the energy sector and zero otherwise. With respect to non-cyclical consumer goods again a dummy is constructed that takes the value of one if the firm that issued the bond operates in the non-cyclical consumer goods sector and a value of zero otherwise.

4.

D

A TA

In order to test whether the bond issuing policy has a significant effect on the spread of a CDS several data is needed to compose a database consisting of the necessary variables. The way these variables are collected and also how they relate to each other will be discussed in this section.

4.1

C

OLLECTION OF DATA

Credit default swap spreads are collected using Datastream. Since credit default swaps are a fairly new derivative, as stated by the international swap and derivative association (www.isda.nl), there is not a large amount of historical data available. Therefore it is decided for this research to stretch a period between 2005 and 2010. To be able to collect a representative dataset there are no geographical limits imposed. Datastream collected a lot of different rates with respect to CDS spreads. For this research however the so called ‘mid-rate’ is used, which is abbreviated by SM. This rate reveals the average of bid and offer quotes. The spread is quoted in basis points of the contract’s notional value. The data in this research is collected on a monthly base for being consistent with the existing literature, for example Collin-Dufresne et al. (2001), Greatrex (2008) and Avramov et al.(2007). Another reason to chose monthly data is to minimize the effect of noise in the data. Eventually a dataset was constructed including a total number of 187 firms with a total of 60 monthly observations. The firms that are included can be found in appendix A.

L.T.M. Lubberman 19 satisfied a set of selected criteria. Bonds with a floating interest rate, convertible bonds, bonds with no clear underlying entity and bonds with missing observations are excluded from this dataset. Out of these collected bonds the maturities are determined and matched to the credit default swaps with the same maturity. For this collection only bonds are used with a maturity between 1 and 10 years, since there are no credit default swaps issued with a maturity longer than 10 year.

To decide on the influence of the maturities of the issued credit default swaps this data is divided in two groups; those with a short maturity and those with a long maturity. As stated by Callen et al. (2007) maturities typically range from one to ten years. Not every maturity is equally often traded. Especially credit default swaps with a maturity of one, two, eight and nine years are quite rare. The absence of these maturities form a natural division and are therefore excluded from this research. The group that consists of credit default swaps with a medium term maturity thus includes maturities between 3 and 7 years. The group with the long term maturities consists of CDS with a maturity of 10 years. This division is comparable to the division used by Collin-Dufesne et al. (2001). Collin-Dufresne et al. (2001) investigates the influence of maturities with respect to bond data. Their short maturity sub-sample includes maturities between three and nine years, the long maturity sub-sample includes maturities for 12 years and longer. Since credit default swaps are not issued with a maturity above 10 years, the division in this research is slightly different than the one used by Collin-Dufresne et al. (2001).

L.T.M. Lubberman 20 Another circumstance under which bonds are issued and that possibly influences the effect of a bond issue on the CDS spread is the sector the firm that issues the bond operates in. The sectors investigated in this research are: energy companies, non-cyclical consumer goods and transport. These sectors are moulded following Thomson Reuters’ indication. These sectors are chosen since they are quite divers and are thus able to capture a large part of the general market movements. The risk-free rate, leverage and volatility are used as control variables. The first control variable is the risk free rate. The risk-free rate has a positive influence on the probability of default. As stated by Ericsson et al. (2004) among others, the higher the risk free rate is, the lower the risk firms take and thus the lower the probability of default. In the literature there has been an extensive discussion about which rate is the most appropriate rate to use. As Hull et al. (2004) describes there are a few different rates that are commonly used. First there is the treasury zero curve rate. This rate is especially used by bond traders and is based on the argument that a bond issued by a government in its own currency has no credit risk, which means that this yield should be equal to the risk free rate. However, it is often forgotten that a government bond tends to be influenced by taxation, liquidity and regulation which clearly affects the yield. For example treasury bonds are often issued to support some regulatory requirements and the interest on these treasury bonds are not taxed on a state level. For these reasons the rate of a treasury bond tends to be too low relative to other low risk bonds. The swap zero curve is more often used by derivative traders. They consider this rate as being more close to the opportunity cost of capital. The swap rate is the credit risk in a series of short term loans from financial institutions to AA borrowers instead of the credit risk of one long term loan from financial institutions to AA borrowers. The important feature of this rate is that the credit risk is very low but not completely risk free and that it is not subject to special tax treatments. Because of these advantages it is decided to use the swap zero curve rate as the risk free rate. Moreover, this rate is also conform existing literature as for example: Alexander and Kaeck (2007), Longstaff et al. (2005), Blanco et al. (2005). This rate is collected with the use of datastream for the corresponding countries and the corresponding maturities.

With respect to the second control variable, leverage is determined conform the formulas used by, among others, Greatrex (2008) and Collin-Dufresne et al. (2001). This formula is as follows:

equity

ue

market val

debt

bookvalue

debt

bookvalue

Leverage

+

=

L.T.M. Lubberman 21 price per share. Datastream is used for collecting this information. The leverage position of a firm influences the probability of default; an increasing leverage position, increases the probability of default of that firm and thus also the credit spread on the bonds of that firm.

The last control variable to be determined is the volatility. As stated by Campbell and Taksler (2003) higher volatility increases default. The volatility is constructed by calculating the firm specific historical 180-day rolling variance of the stock returns preceding the transaction date. This method follows the ones described by Ericsson et al. (2004) and Campbell et al (2004). The stock returns are calculated by collecting the corresponding prices from each firm from Datastream and then transforming them on a logarithmic base into returns.

4.2

D

ESCRIPTIVE STATISTICS

The way the collected data responds to each other is described in this paragraph. Figure 1 shows a picture of how the CDS spread behaves over time.

Date C D S s p re a d i n b a si sp o in ts

Figure 1; This graph shows the mean of the CDS spreads and how they behave over time.

0 50 100 150 200 250 300

IV I II III IV I II III IV I II III IV I II III IV I II III

2006 2007 2008 2009 2010

Mean of CDS

L.T.M. Lubberman 22 middle and could not be assigned to one of the two extremes. With respect to the sectors no observations are excluded.

The mean CDS spreads are quoted in basis points. Compared to the maximum value of the CDS spread and also the standard deviation of the CDS spread, the mean and the median are quite low. The explanation for this phenomenon can be found in the peak of CDS spreads in 2008. Due to the worldwide economic crises, CDS spreads all over the world faced a tremendous growth. This peak is also clearly visible in figure 1. These greatly volatile times thus leaded to this large standard deviation and high maximum values.

To investigate whether CDS spreads of the subgroups are significantly different from each other an ANOVA test is performed. The results can be found in the last column of table 2. An ANOVA test, investigates the assumption that the means of the different groups are all equal. In this research there are three cases where the groups are tested for differences. First there is the maturity case which is divided in two groups to mention bonds with a short maturity and bonds with a long maturity. As can be seen in the table, the ANOVA test is significant which means that there are differences between these groups. This conclusion is conform research of Elton et al. (2001) who state that the spread on bonds with a longer maturity is higher than the spread on bonds with a short maturity. The second case is concerning the business cycle. The defined groups are now periods that face a high business cycle versus periods that face a low business cycle. As can be seen in the table the ANOVA test is significant which means that these groups are significantly different. This is also conform literature of Guha and Hiris (2002) and Moore (1956) that states that credit spreads on bonds are significantly higher during economic downturn than during economic upturn. The last case is with respect to the different sectors. The groups are then non-cyclical consumer goods, transport and energy. The ANOVA test is done for all combinations of 2 sectors. As stated in the table the ANOVA test is again significant for all combinations between these sectors and thus also in this case the groups are significantly different from each other.

L.T.M. Lubberman 23 correlations suggest that there will be no multicollinearity. Multicollinearity, as described by Brooks (2008), exists when the explanatory variables are correlated with each other.

L.T.M. Lubberman 24

#obs Mean Median Max Min Std. Dev .

statistics for the complete sample

CDS spread 11220 97.1333 58.1025 898.5000 2.0000 115.1699

Volatility 11220 0.3674 0.3554 0.8112 0.0027 0.1748

risk-free 11220 0.0362 0.0380 0.0621 0.0042 0.0143

Leverage 11220 0.0778 0.0650 0.5265 0.0062 0.0527

short maturity sub-sample

CDS spread 6480 87.9507 48.7900 898.5000 2.0000 116.4974

Leverage 6480 0.3768 0.3678 0.8112 0.0027 0.1812

risk-free 6480 0.0331 0.0327 0.0621 0.0042 0.0153

Volatility 6480 0.0742 0.0620 0.5265 0.0062 0.0513

Long maturity sub-sample

CDS spread 4740 109.6868 70.3350 802.8000 5.0000 112.1317

Leverage 4740 0.3546 0.3469 0.8112 0.0027 0.1647

risk-free 4740 0.0406 0.0428 0.0594 0.0099 0.0114

Volatility 4740 0.0827 0.0693 0.4621 0.0076 0.0542

Low bu siness cycle scenario

CDS spread 2992 137.8661 87.6500 898.5000 6.4000 142.8101

Leverage 2992 0.3761 0.3714 0.8112 0.0027 0.1778

risk-free 2992 0.0354 0.0395 0.0616 0.0076 0.0123

Volatility 2992 0.0939 0.0773 0.5180 0.0084 0.0611

High business cycle scenario

CDS spread 3366 76.7699 50.8475 658.7300 2.1000 88.4661

Leverage 3366 0.3637 0.3513 0.7653 0.0029 0.1725

risk-free 3366 0.0355 0.0364 0.0570 0.0051 0.0140

Volatility 0.0641 0.0571 0.2248 0.0062 0.0349

consu mer good s scenario

CDS spread 3960 69.3926 49.5175 789.2000 4.5000 69.3754 Leverage 3960 0.2817 0.2816 0.7654 0.0027 0.1420 risk-free 3960 0.0376 0.0395 0.0621 0.0042 0.0137 Volatility 3960 0.0590 0.0509 0.5265 0.0062 0.0407 En erg y scenario CDS spread 5100 135.1913 79.0325 898.5000 2.1000 147.0234 Leverage 5100 0.3908 0.3985 0.8112 0.0476 0.1451 risk-free 5100 0.0398 0.0412 0.0574 0.0097 0.0110 Volatility 5100 0.0947 0.0809 0.4621 0.0077 0.0602

Transport scen ario

CDS spread 2160 58.1321 44.9500 480.2000 2.0000 51.9034 Leverage 2160 0.4695 0.5388 0.7653 0.0031 0.2165 risk-free 2160 0.0252 0.0172 0.0621 0.0042 0.0166 Volatility 2160 0.0722 0.0637 0.2670 0.0084 0.0386 0.00 0.0000 0.0000 0.00 0.00 Anova test on differences A B C D

Table 2; This table shows the summary statistics of the CDS spread, and the variables volatility, risk-free rate and leverage. Panel A shows these statistics without restrictions. Panel B, shows these statistics with respect to the maturity of the underlying bonds issued. Panel C shows the statistics with respect to the business cycle in which the underlying bond was issued and the last panel D shows the statistics of the sectors in which the bond issue took place.

L.T.M. Lubberman 25 l

5.

R

ESU LTS

In this section the research as described before is conducted and results are shown. With the use of eviews several tests have been done. This section starts with looking at the general influence of bond issues on the CDS spread. After that, the described scenarios with respect to cyclical movements and sectors are conducted. The results are published in this section.

5.1

G

ENERAL MODEL

After running the regressions including all the available data and for all maturities, the results are provided in table 3. As can be seen, the variables leverage, risk-free rate and volatility are significant and have the expected signs. The bond issue dummy has a negative sign and is not significant on a 10% level. This sign implies that on average a bond issue lowers the CDS spread change and thus moderates the risk of the underlying entity. One possible explanation is that with a bond issue additional agents are added to look after the risks a firm is taking. By adding agents, the risk preference of a firm is usually reduced. These agents want, as the providers of debt of the firm, to be sure that their money is returned and are usually less willing to take on excessive risk than the providers of equity. However, in this regression this effect is only very weak as can be seen by the small coefficient. When looking at the adjusted R-squared the explanatory power of this model is 6.99%. The explanatory power of the model without the event of bond issue is slightly lower with an amount of 6.95%. This means that the additional variable of a bond issue improves the explanatory power of the model although not to a great extent. The results of this test can be found in appendix D. In general it can be stated that the addition of a bond issue as a variable for explaining the change in credit default swap spread has only a very small influence. The explanatory power increases slightly but the coefficient compared to the other variables is small and not significant.

L.T.M. Lubberman 26 The results when only including credit default swaps with a long maturity can be found in part C of table 3. The signs of the variables have not changed and the coefficients of the control variables also do not show a large change compared to the all maturities case. The coefficient of the bond issue dummy became larger. This implies that the moderating effect with respect to the spread and thus the risk of a CDS is smaller under credit default swaps with a long maturity. This result however is not significant. The adjusted R-squared is slightly higher compared to the case where all maturities are included which means that there is small additional explanatory power by including only credit default swaps with a long maturity. These results are conform research of among others Collin-Dufresne et al. (2001) and Elton et al (2001). Elton et al. (2001) state that there is a general tendency for credit spread on corporate bonds to increase as the maturities lengthens. This result apparently also holds for credit default swaps. The credit default swaps with a short maturity have a more moderated effect on credit spreads than credit default swaps with a long maturity.

L.T.M. Lubberman 27

A

Leverage 181.1551 0.0000 Risk-free -93.0078 0.0000 Volatility 91.7030 0.0000 Bond dummy -0.8720 0.2730 C 0.6264 0.0000 N 11033 Adjusted R-squared 0.0699 1.9751 F-statistic 0.0000

B

Leverage 176.9590 0.0000 Risk-free -95.2709 0.0000 Volatility 104.7625 0.0000 Bond dummy -1.4741 0.1605 C 0.6185 0.0003 N 6372 Adjusted R-squared 0.0761 Durbin-Watson stat 1.9294 F-statistic 0.0000

C

L.T.M. Lubberman 28

5.3

C

YCLICAL ADJUSTMENTS

Table 4 represents the results of the effect of issuing a new bond on the CDS spread adjusted for cyclical movements. The coefficients of the control variables for an issue in the high business cycle as well as an issue in the low business cycle have the expected signs and are significant. With respect to the low business cycle a bond issue has a relatively strong and significant negative effect on the change in the spread of a credit default swap. This effect is not conform the hypothesis based on the research of among others Guha and Hiris (2002), which stated that a cyclical downturn would be accompanied by a higher spread and thus a higher risk. The opposite turned out to be the case. The cyclical downturn leaded to a lower spread during a bond issue. However, when adding the pecking order theory of Donaldson (1961) this result can be explained. This theory states that internal financing is preferred by firms, followed by debt financing. Financing through an equity issue is least preferred. So when the economy finds itself in bad weather and a firm is then still able to receive financing via a debt issue, this might be a positive sign that the market still has enough trust to give the firm debt instead of equity. Thus when a bond is issued during a low business cycle this might be a good sign which reduces the risk on this bond and thus the credit spread on the CDS. The explanatory power of this scenario is slightly higher than the explanatory power of the general case. Also, the explanatory power of the influence of a bond issue during a low business cycle is higher than the explanatory power of the model without the variable of a bond issue during a low business cycle. These results can be found in appendix D. It can be concluded that a bond issue during a low business cycle is a valuable addition to the model explaining the credit default swap spread since the coefficient is significant and the explanatory power is increased. However compared to the coefficients of the control variables the influence of a bond issue is still limited.

L.T.M. Lubberman 29 The explanatory power of this scenario is relatively high with 9.56% compared to the general model and is slightly higher than the explanatory power of this model without the variable of a bond issue which amounts 9.53%. These results can be found in appendix D. It can be stated that the variable of a bond issue during a high business cycle is not a very important variable in explaining the CDS spread since the coefficient of the bond dummy is not significant and the additional explanatory power is only very small.

When looking at credit default swaps with a short maturity the control variables follow the expected signs and are significant at a 5% level. The coefficient of the bond dummy of an issue during the low business cycle is lower than the bond dummy in the general case but not significant any more. The explanatory power is slightly higher. The coefficient of a bond issue during a high business cycle is lower than the coefficient of a bond dummy in the general case but not significant. The explanatory power in this case is lower than in the all maturity case. With respect to including only credit default swaps with a short maturity it can be stated that this does not significantly improves the model. The event of a bond issue is not significant and the explanatory power of the model only shows a marginal to no improvement.

L.T.M. Lubberman 30

A Variable Coefficient Prob. Coefficient Prob.

Leverage 73.8250 0.0000 243.5057 0.0000 Risk-free rate -178.5521 0.0000 -104.3704 0.0000 Volatility 27.2489 0.0243 61.4919 0.0094 Bond dummy 1.0912 0.2871 -4.0063 0.0416 C -0.5333 0.0001 7.5467 0.0000 N 3366 2992 Adjusted R-squared 0.0956 0.0701 2.0744 2.1170 F-statistic 0.0000 0.0000

B Variable Coefficient Prob. Coefficient Prob.

Leverage 74.3881 0.0000 233.1508 0.0000 Risk-free rate -150.9659 0.0000 -111.1202 0.0000 Volatility 30.3618 0.0412 64.6868 0.0380 Bond dummy 0.4741 0.6988 -4.2176 0.1667 C -1.1124 0.0000 8.0194 0.0000 N 1944 1728 Adjusted R-squared 0.0908 0.0704 2.0210 2.1316 F-statistic 0.0000 0.0000

C Variable Coefficient Prob. Coefficient Prob.

Leverage 73.0870 0.0000 260.2273 0.0000 Risk-free rate -218.2251 0.0000 -95.1943 0.0000 Volatility 20.3963 0.3282 56.8746 0.1251 Bond dummy 3.1271 0.1175 -3.9457 0.1236 C 0.2518 0.3046 6.9055 0.0000 N 1422 1264 Adjusted R-squared 0.1068 0.0672 2.1296 2.0949 F-statistic 0.0000 0.0000 Durbin-Watson stat

Table 4; model with the effect measured of the business cycle of the bond issued.The regression is in the following form: ΔCDS=α+βΔrf+βΔlev+βΔvol+γbond D+ε. Where rf stands for risk free rate, lev means leverage and vol means volatility. Part A includes CDS's with all maturities, part B only CDS's with a maturity between 3 and 7 year and part C only CDS's with maturities of 10 years.

L.T.M. Lubberman 31

5.4

S

ECTOR ADJUSTMENTS

This section looks at the effect of a division in sectors on the CDS spread change when a bond issue takes place. The results can be found in table 5. With respect to all sectors, the control variables are significant and follow the expected sign. With respect to the effect of a bond issue, the sector energy has the strongest negative sign and is as the only one significant. The hypothesis stated that with energy being the least capital intensive sector bond issues would likely cause the smallest change in spread. The results seem to be conform theory. The large coefficient suppresses the changes in CDS spread and thus the risk will be lower. However, as can be seen in the table, the adjusted R-square is very low compared to the general case in table 3. Compared to the outcomes of the model without a bond issue as an additional variable, the explanatory power only slightly increases. These results can be found in appendix D. Thus although an issue in the energy sector does significantly influences the CDS spread, this influence is so small that it probably has a limited effect.

With respect to transport it was expected that this section, as being the most capital intensive, would be the most sensitive to a bond issue. These results are conform theory. The small but positive coefficient influences the CDS spread change by making this larger. However, one must notice that this effect is not significant. The explanatory power of this model is again quite low and shows only a very small increase compared to the model without a bond issue as additional variable. Since the coefficient of the bond issue is not significant and the explanatory power only slightly increases by adding the additional variable of a bond issue, it can be stated that in the transport sector a bond issue is not a very important determinant in explaining an additional portion of the CDS spread.

The sector non-cyclical consumer goods was hypothesized to be in the middle with respect to capital intensiveness of the sectors. With a coefficient that lies in between these values, this hypothesis holds. However, also this sector does not have a significant influence on the change of a credit default swap spread. This sector shows a relatively high R-squared compared to the general case and thus a relatively high part of the risk is explained. However, comparing this model to the model without bond issue as an additional variable shows even a slightly higher explanatory power. This means that although with respect to credit default swaps of firms operating in the non-cyclical consumer goods sector the model shows a relatively good fit, this fit is not improved by adding the additional variable of a bond issue. The insignificance of the coefficient of the bond issue dummy strengthens this conclusion.

L.T.M. Lubberman 32 energy sector and the non-cyclical consumer goods sector is lower than these coefficients in the general case. The coefficient of a bond issue in the transport sector is higher. The explanatory power increases in the non-cyclical consumer goods sector and the transport sector, but shows a small decrease in the energy sector. With respect to including only credit default swaps with a short maturity it can be stated that this does not lead to an improved fit of the model. The coefficients of the bond issue dummy do not show a significant influence in explaining an additional part of the observed CDS spread change and also the explanatory power of the model does not show an important improvement.

L.T.M. Lubberman 33

A Variable Coefficient Prob. Coefficient Prob. Coefficient Prob.

Leverage 200.7901 0.0000 195.9158 0.0000 108.9714 0.0000 Risk-free rate -75.7627 0.0000 -98.4642 0.0000 -19.5327 0.0000 Volatility 96.2321 0.0000 93.3333 0.0000 66.6792 0.0006 Bond dummy -0.4658 0.6752 -4.3811 0.0384 0.2021 0.8733 C 0.1429 0.4271 1.1484 0.0005 0.3491 0.1859 N 3894 5015 2124 Adjusted R-squared 0.1081 0.0445 0.0682 Durbin-Watson stat 1.8936 2.0643 1.9553 F-statistic 0.0000 0.0000 0.0000

B Variable Coefficient Prob. Coefficient Prob. Coefficient Prob.

Leverage 218.5557 0.0000 163.6360 0.0000 97.8580 0.0000 Risk-free rate -82.9889 0.0000 -10.3179 0.0000 -215.9952 0.0000 Volatility 98.2061 0.0000 92.2737 0.0000 114.2347 0.0000 Bond dummy -1.1910 0.4039 -6.1444 0.0284 0.6257 0.7230 C 0.2662 0.2224 1.1413 0.0048 0.4266 0.2091 N 2714 2537 1121 Adjusted R-squared 0.1124 0.0439 0.0910 Durbin-Watson stat 1.8620 2.0204 1.9399 F-statistic 0.0000 0.0000 0.0000

C Variable Coefficient Prob. Coefficient Prob. Coefficient Prob.

Leverage 171.0993 0.0000 251.4942 0.0000 127.6073 0.0000 Risk-free rate -61.0162 0.0000 -92.8130 0.0003 -189.0420 0.0000 Volatility 101.8101 0.0000 94.8227 0.0006 -6.2720 0.8409 Bond dummy 1.0268 0.5661 -2.5190 0.4376 -0.0414 0.9821 C -0.1536 0.6342 1.1129 0.0530 0.2711 0.5335 N 1180 2478 1003 Adjusted R-squared 0.1050 0.0476 0.0865 Durbin-Watson stat 1.9734 2.1182 1.9819 F-statistic 0.0000 0.0000 0.0000 Long maturities

Consumer goods Energy Transport

Short maturities

Table 5; model with the effect measured of the sector in whic the bond issued.The regression is in the following form:

ΔCDS=α+βΔrf+βΔlev+βΔvol+γbond D+ε. Where rf stands for risk free rate, lev means leverage and vol means volatility. Part A includes CDS's with all maturities, part B only CDS's with a maturity between 3 and 7 year and part C only CDS's with maturities of 10 years.

Consumer goods Energy Transport

L.T.M. Lubberman 34

5.5

I

NFLUENCE OF THE VOLATILITY

To get further information about the robustness of the results of the regressions performed above an additional regression analysis is performed. In this analysis the sample data is divided in two sub-samples on a time series base. The first sub-sample covers the first two years of the sample, and the second sub-sample covers the last three years of the sample. This means that a division is made at the beginning of 2008. To appoint this moment as the splitting point of these sub-samples is an important implication of this robustness check. As can be seen in figure 1 the changes in CDS spread became much more volatile since the beginning of 2008. This phenomenon can be appointed to the world wide outburst of the credit crisis. The difference in descriptive statistics is shown in appendix E. The table in the appendix and figure 1 make it obvious that the times has changed.

The results of these regressions are remarkable and can be found in appendix E. With respect to the general model without including scenarios the first sub-sample show that the control variables Volatility and leverage are not significant anymore and also their coefficients show a large change. In this case a bond issue however, is a significant determinant of the CDS spread. However, the explanatory power of this regression is so low that not much can be concluded from these results. In the second sub-section, which is the more volatile period, the control variables are significant again and the model even gained in explanatory power compared to the all months case. The event of a bond issue is not a significant variable to influence the CDS spread but the explanatory power of the test has increased compared to the general model in table 3. The fit of this model has thus improved in the more volatile period.

L.T.M. Lubberman 35

6.

C

ONC LUS ION

This study investigates whether a bond issue influences the spread of a credit default swap. By determining what factors influence the credit default swap spread, this spread can be more accurately predicted and a better insight is provided in factors that influence corporate risk. Former research of among others Ericsson et al. (2004) already determined the leverage position of the underlying firm, equity volatility and the risk-free rate as important variables for determining the CDS spread. To investigate whether a bond issue and the maturity of the CDS are valuable additions to this set of variables a database is constructed including 187 firms over a time period of 60 months. For this research the collected data is structured as panel data for being able to test simultaneously on a cross-sectional dimension as well as on a time series base. Since the influence a bond issue has, may depend on the circumstances under which the bond is issued, a few of these circumstances are investigated in this research to mention the business cycle the firm operates in at the moment of the issue and the capital intensiveness of the sector the firm operates in. Apart from these circumstances, also the influence of the maturity of the credit default swaps is investigated.

With respect to the influence of the maturity of a bond issue, the hypothesis is based on the research of Elton et al. (2001) and states that a shorter maturity of a CDS issue leads to a lower credit spread. Although this effect is visible in the coefficients of the bond issues under all the different circumstances, the results are not significant.

L.T.M. Lubberman 36 With respect to the influence of the sector the firm operates in, it is hypothesised according among others by Belkaoui (1980) that bonds issued in a highly capital intensive sector increases the observed spreads of credit default swaps. The results have provided evidence that an issue in the energy sector (the least capital intense sector) has a significant impact on the change in CDS spread. This result is conform theory that states that least capital intensive sectors face lower credit risk and thus a lower CDS spread. However, this regression had an exceptionally low adjusted R-squared, so it is hard to draw conclusions solely based on this result. The other two sectors, transport and non-cyclical consumer goods did not show significant results.

Although this research is carefully constructed there are some limitations. One limitation is that to be conclusive a complete view of the market is needed. Due to the selection of only three sectors, the market might not be completely covered. Weather these results also hold with other sectors need further research. Another limitation is that this research is one of the first to use panel data techniques. Although this data structure suits this data exceptionally well, it is not used much yet in other literature. This makes it harder to give a good comparison between the outcomes of this research and former researches. Further research in the use of data panel technique in identifying determinants with respect to credit default swaps might thus be useful.

In the years before the outburst of the credit crisis the popularity of credit default swaps experienced an enormous growth. This can reasonably be explained by increased uncertainty about stock markets, housing prices and the liquidity of banks. One thing this crises made visible again is that the market has many imperfections. Especially financial intermediaries disturbed the market by becoming ‘too big to fail’. Since not only financial intermediaries but also many corporations are too big to fail many professionals these days think that when firms are not longer allowed to get in default, derivatives in the form of a CDS might become useless and will soon be outdated. However, one of the characteristics of a CDS is that the definition of default can be set by the seller. If the definition of default is more commonly put to a wider set of events (for example not only the inability to make the regular payments but also a decrease of their credit rating can be specified as a default event), credit default swaps might preserve this popularity and remains its importance in understanding what factors influence the change in credit spread of a CDS.

L.T.M. Lubberman 38

R

EFERENC ES

Abid, F. and N. Naifar., 2006, The determinants of credit default swap rates: an explanatory study, International Journal of Theoretical and Applied Finance, 9, 23-42.

Alexander, C. and A. Kaeck, 2007, Regime dependant determinants of credit default swap spreads, Journal of Banking and Finance, 32, 1008-1021.

Avramov, D., G. Jostova and A. Philipov, 2007, Understanding changes in corporate credit spread, Financial Analyst Journal, 63, 90-105.

Baxter, M. and R. King, 1999, Measuring business cycles: approximate band-ass filters for economic time series, Review of Economics and Statistics, 81, 575-593.

Belkaoui, A., 1980, Industrial bond rating: a new look, Financial Management, 9, 44-51.

Benkert, C., 2004, Explaining credit default swap premia, Journal of Futures Markets, 24, 71–92. Berndt A., R. Douglas, D. Duffie, M. Ferguson, and D. Schranz, 2004, Measuring default risk premia from default swap rates and EDFs, Working Paper. Cornell university.

Black, K. and M. Scholes, 1973, The pricing of options and corporate liabilities, Journal of Political Economy,7, 637-654.

Blanco, R., S. Brennan and I. Marsh, 2003, An empirical analysis of the dynamic relationship between investment grade bonds and credit default swaps. Working Paper, Bank of England.

Callen, J., J. Livnat and D. Segal, 2009, The impact of earnings on the pricing of credit default swaps, The Accounting Review, 84, 1363-1394.

Campbell, J., and G. Taksler, 2003, equity volatility and corporate bond yields, Journal of Finance, 58, 2321-2349.

Chen, R., X. Cheng, F. Fabozzi, and B. Liu, 2008, an explicit, multi-factor credit default swap pricing model with correlated factors, Journal of Financial and Quantitative Analysis, 43, 123-160. Collin-Dufresne, P., R.Goldstein and J. Martin, 2001, The determinants of credit spread changes, Journal of Finance, 56, 2177–2207.

Duffie, D. and K. Singleton, 1999, Modeling term structures of defaultable bonds, Review of Financial Studies, 12, 687-720.

Elton, E., M. Gruber, D. Agrawal and C. Mann, 2001, Explaining the rate spread on corporate bonds, Journal of Finance, 56, 247-277.