Credit risk in Islamic joint venture bond

Accepted Manuscript Title: Credit Risk in Islamic Joint Venture Bond Author: Saad Azmat Michael Skully Kym Brown PII: DOI: Reference:

S0167-2681(14)00090-0 http://dx.doi.org/doi:10.1016/j.jebo.2014.03.020 JEBO 3332

To appear in:

Journal

Received date: Revised date: Accepted date:

9-11-2012 7-3-2014 19-3-2014

of

Economic

Behavior

&

Organization

Please cite this article as: Azmat, S., Skully, M., Brown, K.,Credit Risk in Islamic Joint Venture Bond, Journal of Economic Behavior and Organization (2014), http://dx.doi.org/10.1016/j.jebo.2014.03.020 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Credit Risk in Islamic Joint Venture Bond *

an

us

Professor Michael Skully Department of Accounting and Finance Monash University P.O. Box 197 Caulfield East Victoria 3145 Australia Ph: 61 3 9903 2407; Fax: 61 3 9903 2422; Email: [email protected]

cr

ip t

Saad Azmat Assistant Professor Suleman Dawood School of Business Lahore University of Management Sciences D.H.A, Lahore Cantt, 54792 Pakistan Ph: 92 42 111 11 5867; Email: [email protected]

Ac ce p

te

d

M

Dr Kym Brown (corresponding author) Department of Accounting and Finance Monash University P.O. Box 197 Caulfield East Victoria 3145 Australia Ph: 61 3 9903 1053; Fax: 61 3 9903 2422; Email: [email protected]

1

Page 1 of 60

Credit Risk in Islamic Joint Venture Bonds

ip t

This paper focuses on Islamic joint venture (IJV) bonds and examines whether conventional structural credit risk models capture Islamic bonds’ underlying risk. Their various extensions

cr

have first been adjusted for the IJV bonds’ unique characteristics and then tested through simulations to identify any appraisal issues. The models are then used to assess data from 52

us

Malaysian Islamic bond issuers. We find that conventional structural models and their Islamic

d

JEL: G15, G24

M

Keywords: Credit risk; Islamic bonds

an

extensions misevaluate IJV bonds and so afford them lower credit ratings.

Ac ce p

te

Acknowledgements: We would sincerely like to thank the anonymous reviewers and the editor for their suggestions to improve our paper and as well as participants at the JEBO Islamic Finance Conference, 29th September – 1st October 2012, Birmingham.

2

Page 2 of 60

Credit Risk in Islamic Joint Venture Bonds

ip t

1. Introduction

The Islamic finance industry has grown rapidly over the last few decades to total some USD

cr

$1.54 trillion in financial assets by the end of 2012 (EY, 2014). Of these, USD $130 billion were

us

accounted for by sukuk issues - financial instruments often known as Islamic bonds. This growth has been hampered by a shortage of specially trained and experienced practitioners in Islamic

an

finance. So in the short run, the industry has instead hired conventional bankers as well as applied conventional finance techniques across much of their operations. While expedient, this

M

solution has not been without its costs. Some Islamic instruments now look all too similar to conventional ones and likewise some techniques have been applied with only a modest

te

d

consideration of true risk and reward positions that Islamic finance affords.

Ac ce p

This is particularly so for the sukuk where as with other financial instruments, risk and reward are the key determinants in their valuation and hence attractiveness. Some behavioural studies suggest that investors have a much stronger aversion for uncertainty and loss than implied by expected utility theory, hence, a higher weight is assigned to risky outcomes in their utility (Kahneman and Tversky, 1979; Benatzi and Thaler, 1995). Misevaluation of risk, therefore, can result in investors experiencing artificially high or low levels of utility from investing in certain instruments. In conventional finance, such misevaluations have led to mispricing, bubbles and inefficient distribution of financial resources (Brunnermeier, 2009; Bolton, Freixas and Shapiro, 2012; Acharya and Naqvi, 2012). For risk measurements of debt instruments, credit risk models 3

Page 3 of 60

play an important role. Their validity and assumptions have been continuously updated for new and complex financial instruments (Duffie and Singleton, 2003; Lando, 2004). The journey from the initial Merton (1974) credit risk model to the practical Finger et al. (2002) CreditGrades

ip t

model as well as to the more complex reduced form models (Duffie and Singleton, 2003) took the conventional finance industry several decades. In comparison, Islamic finance has had no

cr

such experience and in most cases still uses conventional credit risk models (Usmani, 20071;

us

RAM ratings, 2011), perhaps quite inappropriately, to capture the credit risk of Islamic bonds.

an

Islamic bonds in general and Islamic joint venture (IJV) bonds 2 in particular possess some unique characteristics, e.g. IJV bonds have a more pronounced equity link. Conventional credit

M

risk models, which are designed to capture the risk of debt like structures seem to be incapable of

d

capturing the risk of IJV type structures and might even assign them higher default probabilities

te

(lower survival probabilities) compared to conventional debt bonds. Researchers, however, have seemingly not challenged the validity nor questioned the ability of conventional credit risk

Ac ce p

models to identify their underlying risk differences. This research gap motivates the following research question: Can conventional credit risk models be used to assess credit risk in IJV bonds?

This paper adjusts, the Merton (1974), first passage and CreditGrades model for the IJV bonds’ unique characteristics and tests through simulations to identify if these models have a misevaluation against them. The adjusted models are then used to test data from 52 Malaysian 1

The prospectus from the sukuk issuance by Sitara Chemical Musharakah Term Finance Certificate Prospectus stated that conventional credit models have been used in their analysis of various sukuk products. As conventional finance professionals simply translate sukuk as Islamic bonds, IJV related sukuk are promoted as simply as a different type of bond even though they have a much stronger equity nature. 2 IJV bonds are more formally known as Musharkah Sukuk. 4

Page 4 of 60

Islamic bond issuers. The results suggest that these models and their Islamic extensions have a bias against equity type IJV bonds resulting in lower credit ratings. This is because current credit risk models focus on the principal’s repayment abilities. In contrast, IJV bonds’ return can be

ip t

greater (or lower) than the principal. The results also imply that this bias cannot be removed by merely adjusting conventional models for IJV characteristics. So perhaps credit rating agencies

us

cr

should treat IJV bonds simply as equity and measure their risk using equity risk tools.

an

This paper contributes to two distinct strands of literature. Firstly, in the credit risk modelling literature (Duffie and Singleton, 2003) it suggests that conventional models might be

M

misevaluating the risk of IJV bonds. It also traces this misevaluation to a tacit model assumption, a sole focus of safekeeping the principal. The principal’s safekeeping and other conventional

d

model assumptions should therefore be closely examined to ensure they reflect the nature of

te

different Islamic instruments for both Islamic capital markets and Islamic banks. Secondly, the

Ac ce p

paper adds to the debate over whether Islamic instruments little originality different from conventional instruments (Chong and Liu, 2009; Khan, 2010) by suggesting that the reliance on conventional credit risk models could explain the poor utilisation of IJV bonds.

The remainder of this paper is structured as follows. Section 2 discusses the unique features of IJV bonds. Section 3 provides the literature review. The methodology and data is discussed in Section 4. The results are presented in Section 5 and the paper ends with the summary in Section 6.

5

Page 5 of 60

2. IJV Bonds Islamic capital markets use alternatives to conventional finance that are free from interest (riba),

ip t

uncertainty (gharar) and gambling (maysir). As detailed in Table 1, the most commonly used structures include IJV (Musharakah/Mudarabah) bonds, secured against real asset (Ijarah) bonds

us

cr

and debt (Murabahah) bonds (Ayub, 2007). The focus in this paper is on IJV bonds.

an

(Insert Table 1 about here)

IJV bonds (Musharakah/Mudarabah sukuk) are unique instruments with characteristics of both

M

debt and equity. Their holders effectively become part owners of a particular project. If the

d

project is successful, IJV bond holders receive a periodic share of the profits, in an attempt to

te

safeguard the principal, and similar to debt characteristics. IJV bond holders also bear any of its loss according to their investment proportion. Unlike ordinary equity shares, IJV bonds have a

Ac ce p

maturity date at which time unlike normal debt or preference shares, the issuer can, and usually does, buy them back from the investor at their then market price rather than face value (Usmani, 2002). Their equity type structure implies that, given no arbitrage, there should be a direct relationship between the market price of a firm’s IJV bonds and its ordinary shares. This is because IJV bond holders, like shareholders, are entitled to both profits and losses.3 Moreover, in case of bankruptcy, IJV bond holders rank equally with shareholders but not higher as would be the case for debt bonds. This differentiates IJV default from a normal debt default. In a

3

If IJV bonds are used to finance a particular project then this profit and loss sharing is restricted to that project only. 6

Page 6 of 60

conventional bond, a default occurs when the debtor fails to repay the principal or the interest payment. In an IJV bond, like an ordinary share, neither the principal nor the return is guaranteed.

ip t

Therefore, a traditional default event does not exist for an IJV bond.

cr

3. Literature Review

In a world with rational actors, perfect information and efficient markets, an instrument’s

us

underlying risk4 would always be reflected in its price. There would be little need for rating

an

agencies or sophisticated risk measurement tools. Without these ideal assumptions though, an understanding of risk and tools to measure it becomes important. This section begins with

M

discussing the importance of risk in investor’s financial decision and the impact of risk possible misevaluation. Then it examines the nature of credit risk in conventional bonds and how

d

structural models such as the Merton (1974) model, first-passage models and CreditGrades

Ac ce p

and develop the hypotheses.

te

model are used to measure credit risk. Finally, we examine the nature of credit risk in IJV bonds

3.1 Risk, Decision Making and Misevaluation Behavioural studies starting with Kahnemen and Trvesky (1979) suggest that uncertainty has a much more important role than implied by expected utility theory. Risky outcomes therefore are assigned a higher weight in investment decisions. One strand of research argues that investors assign higher weights to loss making situations compared to the outcomes representing gains (Bentazi and Thaler, 1995). Given the importance of risk in the investor’s decision making, a

4

While this paper concentrates solely on credit risk as do credit risk models, there are of course a range of other risks to consider such as market risk, liquidity risk, fraud and Shariah compliance risk. 7

Page 7 of 60

misevaluation of risk can have significant repercussions including mispricing, bubbles and inefficient distribution of financial resources (Bolton, Freixas and Shapiro, 2012; Acharya and Naqvi, 2012). In the recent sub-prime financial crisis, risk misevaluation of certain financial

ip t

instruments was considered a primary cause (Brunnermeier, 2009). In such an environment, the rating agencies and the financial models they employ become important, particularly in the bond

Evaluation of Credit Risk

us

3.2

cr

market. The next section overviews the credit risk models used in the literature.

an

The credit risk modelling literature can be divided into three broad categories: traditional methods, structural models and intensity or reduced form models. The traditional methods

M

involve a non-modelling approach and mainly use historical data to infer default probability (Ammann, 2001). Structural models compare firm value with that of its debt (Duffie and

d

Singleton, 2003). In contrast, intensity or reduced form models, rather than linking default to a

te

firm’s financials, adopt a default process.5 For reasons mentioned later in the section, we focus

Ac ce p

on structural models. Firm value is comprised of an equity and debt component and their comparison establishes a strong link between credit and equity markets (Finger et al., 2002). Structural models in turn can be divided into two categories: firm value models and first-passage models. The firm value models, such as the Merton model, compare the value of the firm with that of its debt only at a typical maturity date and so only experience default at maturity. First passage models, in contrast, assume that default can occur any time during the bond’s life.

5

The default process is a one jump process where it can go from no-default to default with a particular probability governed by the intensity of default. Some of the intensity models have been developed by Jarrow and Turnbull (1995), Jarrow et al. (1997), Duffie and Singleton (1999). 8

Page 8 of 60

The initial Merton’s (1974) model was developed to price bonds when default likelihood was significant. It assumes that equity holders are owners of the firm’s assets, who buy a put option from the debt holders. In the case of default, equity holders could exercise the put option and

ip t

extinguish their debt. As default can occur only at maturity, the Merton (1974) model calculates

cr

the survival probability by comparing the asset value at maturity with that of its debt.

us

Though the Merton (1974) model offers an intriguing approach to risk measurement, its maturity dependent default assumption of maturity dependent default is unrealistic in practice. This

an

problem is remedied by the first passage models which incorporate the assumption of default occurring anytime during bond maturity (Duffie and Singleton, 2003). Default in the first

M

passage model occurs when the firm’s asset value falls below a typical default boundary 6

d

irrespective of the maturity date. Most first passage models include a safety covenant, which give

te

the asset holders the right to enforce bankruptcy once the asset value goes below the default boundary Moreover, once the default boundary is reached, the bond holders can claim the firm’s

Ac ce p

assets (Black and Cox, 1976). In the simplistic first passage model, the debt level is taken to be the default boundary, implying that on default the bond holder’s principal debt would keep intact. This assumption is unrealistic as it implies a perfect recovery rate. In reality, the default boundary should be dependent on multiple factors including the time to maturity 7 and the

6

For a simplistic case the debt level can be taken as the default boundary. In this simplistic first passage model, therefore, default occurs once the firm value falls below the debt level anytime during the maturity of the bond. 7 Term to maturity was first incorporated in the default by Black and Cox (1976). They assume an exponential default boundary. In their model as it nears maturity, the default boundary and the amount the bond holder receives on default increases. A slight variant of the default boundary is used by Brennan and Schwartz (1980) who assume a constant default boundary independent of the time to maturity. This implies that the amount the bond holders receive on default is independent of maturity. 9

Page 9 of 60

recovery rate8 (Duffie and Singleton, 2003; Meissner, 2005). A first passage model that makes changes to the default by incorporating an uncertain recovery rate is CreditGrades.

ip t

CreditGrades is a first passage model jointly developed by four leading credit market institutions: The RiskMetrics Group, JP Morgan, Goldman Sachs and Deutsche Bank. It attempts to rectify

cr

the unrealistically low short-term spread generated by first passage models (Finger et al., 2002).9

us

The CreditGrades model achieves this by assuming an uncertain recovery rate. This uncertainty is reflected in the default boundary by causing the firm value to hit the default boundary

M

an

unexpectedly.

3.3 Credit Risk in IJV Bonds

d

IJV bonds have much in common with equity. Foremost is their shared risk and return

te

characteristics where bond holders are entitled to a proportion of the venture’s profit as well as

Ac ce p

are liable for its downward risk (Usmani, 2002; Howladar, 2006). This introduces uncertainty in IJV redemption value at maturity, as it can be less or more than the IJV bond’s initial principal.

This uncertainty, therefore, has to be incorporated first in the default boundary of Merton (1974), first passage and CreditGrades model to produce IJV extensions. A greater challenge however, is 8

Recovery rate is another important determinant of the default boundary. It is incorporated by Longstaff and Schwartz (1995) who use a constant default boundary k and recovery rate w. Both k and w are determined exogenously. They also assume that the senior debt would have higher recovery rate than junior debt. One shortcoming of their model is that the payout on default can be at times greater than the firm’s value. This is corrected by Briys and deVarenne (1997) who ensure that the payout on default cannot exceed the asset value. They do this by using a default boundary which is dependent on the price of the risk free bond and face value of the risky bond. Their model also assigns different recovery rates for default occurring pre-maturity and at maturity. 9 The reason is that asset value cannot reach the default boundary quick enough through pure diffusion. 10

Page 10 of 60

that structural credit risk models focus on the probability of returning only the principal. In contrast, an IJV bond holder might not only be interested in the principal but also the probability of the different profitability states that the bond might generate. By simply incorporating the IJV

ip t

redemption value uncertainty in the default boundary and ignoring these other profitability states, these IJV extensions will assign artificially low survival probabilities to IJV bonds. This can be

cr

illustrated by comparing the survival probabilities generated by the original structural models

us

with those of their IJV extension and noting the differences. We aim to therefore test:

an

H1: The IJV extensions of structural bond credit risk models generate significantly lower

M

survival probabilities than the original conventional bond credit risk models.

d

The effectiveness of structural models’ IJV extensions can also be tested by comparing how they

te

rank survival probabilities of firms in terms of their credit risk (Finger et al., 2002). A high positive association10 between the IJV extensions’ rankings and those of the original models (or

Ac ce p

credit ratings11) reflect their sound working. In contrast, a negative association might reflect an IJV misevaluation. For example, the CreditGrades model might rank a firm in the top 10% of the least risky bond issuers but CreditGrades IJV extension might rank it in the bottom 10%. This leads to testing:

H2: There is a negative association between bond credit risk rankings generated by the structural models’ IJV extensions and their original bond credit ratings and rankings. 10

In Section 4.6.3, two measures of association are used: correlation coefficient and Kendall’s tau. Most Islamic bonds are rated using conventional credit risk models (Usmani, 2007), therefore, rankings generated using issuer’s credit ratings can also be used to compare with the results of structural bond models’ IJV extensions.

11

11

Page 11 of 60

4. Methodology and Data

ip t

This section discusses the IJV extension methodology for the structural models i.e. the Merton (1974), first passage and CreditGrades models. The credit risk nature of IJV bonds is analysed

cr

first followed by an investigation of their market value. Their unique features are then

an

credit score methodology to test these models is examined.

us

incorporated in three models respectively to develop their IJV extensions. Finally the data and

M

4.1 Credit Risk Nature of IJV Bonds

Credit risk models are built to capture bonds’ probability of default (or survival) (see Equation 3).

d

Typically with conventional bonds, default occurs when the issuer fails to pay the principal or

te

interest (Duffie and Singleton, 2003). IJV bonds are like ordinary shares however, and represent

Ac ce p

ownership in the venture and do not guarantee the principal. Moreover, at maturity the bond holder receives the market value of IJV bonds which can be greater or less than the principal. Traditional default does not exist in IJV bonds. An important assumption behind the derivations of the structural models is that the firm’s equity is a call option on the asset value with a strike price (as shown in Equation 3, with the Merton (1974) model the strike price is equal to the principal amount). This is true for debt in practice but not IJV bonds. This is, however, a useful theoretical characterization which allows us to compare survival probabilities. It is, therefore, used for developing IJV extensions by incorporating an important reservation that in case of IJV default the principal is not guaranteed and loss is shared. Equation 6 captures these features.

12

Page 12 of 60

The structural model’s IJV extensions must assume (rather inappropriately) that IJV bond holders are only concerned with the safety of their principal. Moreover, they would dissolve the venture to obtain the market value of their ownership on the fear of loss. This assumption,

ip t

however, ignores the high positive returns that IJV bonds may generate, hence, creating a misevaluation. A principal focused IJV extension would simply incorporate the uncertainty about

cr

the IJV redemption (or market) value in to the original structural model’s default boundary.12

us

The model assumes that the actual recovery would depend on the market price of the firm at default and other costs of financial distress. Note the proofs of the credit risk models IJV

M

an

extensions are given in Appendix A.

4.2 IJV Market Value

d

IJV bonds have a number of features common with equity, particularly in the profit and loss

and the stock price of the issuing firm (

Ac ce p

market value of IJV bond (

te

sharing with the shareholders. Arbitrage should, therefore, ensure a relationship between the , captured by:

(1)

This relationship between IJV market value and stock price could depend upon several factors including IJV bonds’ partial ownership, lack of shareholder privileges and liquidity in the secondary markets. Partial ownership is where it finances individual projects where the IJV

12

A typical IJV bond relationship allows the IJV bond holder to dissolve, liquidate or sell the venture and claim part of its initial principal (these caveats can be included in the venture). 13

Page 13 of 60

owners are limited to that project only. In contrast, ordinary shares own the whole firm. Their pricing relationship, therefore, would depend on the proportional IJV project value compared to the total value of the business. Lack of shareholders’ privileges available to IJV bond holders

ip t

(like voting rights) might also result in a lower price compared to the voting shares. Finally poor secondary markets may also affect arbitrage opportunities between IJV and stock prices. A

cr

highly illiquid IJV bond market for example could result in IJV being lower than the firm’s stock

us

price. This illiquidity effect is a function of term to maturity and should dissipate completely at maturity when the issuer promises to redeem these bonds at their market value. Given the similar

an

characteristics of IJV bonds and stock value, arbitrage would ensure that the IJV bond price Mt in the bond market would be a function of the stock price of the issuing firm (St) in the equity

M

market (Equation 1 and 2). The differences discussed here are captured in the IJV and stock price

Where

(2)

Ac ce p

te

d

relationship such that:

represents factors such as the partial ownership of IJV bond holders and the differences

in share privileges, while ‘a’ measures factors such as bond market liquidity. Bonds that are perfectly liquid ‘a’ will be assigned a value of zero, while higher values reflect increasing illiquidity. Equation 2 shows that the illiquidity effect dissipates as the bond approaches maturity and is neutralised at maturity date ‘T’.

Under IJV bond financing the firm value V = S (equity) + M (market value of the IJV bond, rather than debt, D) and on default the holders get their share of M rather than firm value (V). 14

Page 14 of 60

4.3 Merton Model and IJV Extension The IJV credit risk and market value features are extended using the Merton (1974) model to create an IJV extension. In the Merton (1974) model with a certain recovery rate, default occurs

ip t

should the outstanding debt (D) exceed the firm’s value (VT) at maturity. Equation 4 gives the probability of survival P(Z) generated by the model. Here, σ2 is the standard deviation of stock

(3) (4)

(5)

d

M

an

us

cr

prices and σv2 is the standard deviation of firm value.

te

The IJV extension of the Merton (1974) model assumes that IJV bond holders would receive the

Ac ce p

market value of their bonds rather than simply their principal at maturity. The IJV extension, therefore, captures the probability of its value at maturity being greater than at issuance and is represented by:

Where

(6)

is the market value of IJV bonds at the issuance date. The no-arbitrage condition13

ensures that the IJV bonds market value should reflect the stock price as represented by Equation 13

No arbitrage condition means that the all opportunities for arbitrage have been exhausted. 15

Page 15 of 60

2. Substituting Equation 2 in Equation 6 leads to Equation 7. This can be solved to generate the survival probability of Merton (1974) model’s IJV extension as shown in Equation 8.

ip t

(7)

an

4.4 First Passage Model and IJV Extension

us

cr

(8)

In a first passage model, default can occur any time during bond maturity. Equation 9 shows the

(9)

Ac ce p

te

d

M

survival probability for the first passage model.

The IJV extension of the first passage model shows “default” when the IJV bond’s market value (Mt) falls below its issue value (M0) at any time during the bond’s life. This is captured by:

(10)

16

Page 16 of 60

ip t

(11)

(12)

us

cr

For simplicity assume that

It is also assumed that on default the illiquidity premium disappears, hence, T-t = 0 and

.14

an

Equation 11 above can be solved to generate the survival probability of the first passage IJV

(13)

Ac ce p

te

d

M

extension model as shown by Equation 13.

4.5 CreditGrades Model and IJV Extension A distinguishing feature of the CreditGrades model is how it captures the uncertainty in the recovery rate and hence the default boundary (Finger et al., 2002). Equation 17 shows the CreditGrades survival probability. 14

It is for simplicity only and does not change the outcomes of the results. 17

Page 17 of 60

ip t

(14)

(16)

Ac ce p

te

d

M

an

us

cr

(15)

(17)

(18)

(19)

18

Page 18 of 60

The CreditGrades IJV extension incorporates the uncertainty assumption in the recovery rate by assuming that the amount IJV bond holders receive upon default (when the venture is dissolved) can differ from market price given by Equation 2. This uncertainty could be due to financial

Ac ce p

te

d

M

an

us

cr

20 and 21 capture the survival probability of CreditGrades IJV extension.

ip t

distress costs such as litigation or price discounts during the sale of illiquid real assets. Equations

(20) (21) (22) (23) (24) (25)

(26)

(27)

For all the simulated models, survival probabilities are generated against term to maturity and stock volatility. In the standard cases the firms are assumed to have a low debt ratio while operating in a highly liquid IJV bond market (with a low value of a = 0.01).

4.6 Credit Rating by Issuers vis-à-vis Ratings by Structural Models

19

Page 19 of 60

This section now applies our models to real data to compare our result with their actual credit ratings and rankings. Our data consists of 52 corporate Malaysian Islamic bonds issuers for the period 2002 to 2010. The sample consists of 4 SARA bond issuers, 7 IJV bond issuers and 41

ip t

Islamic debt bond issuers. An examination of the IJV bond structures, however, suggests that most of them are actually structured in a debt like manner.15 This is not surprising given the

cr

warnings by a number of Shariah advisors and AAOIFI (AAOIF, 2008). The issuer’s credit

us

ratings and term to maturity are taken from the IFIS data base. Daily data for firm specific variables, such as share price and volatility, is taken from Datastream. Yearly data for debt per

an

share is taken from the Compustat. The sample contains only corporate bond issuers while data for banks and other financial institutions are excluded as some of their financial ratios are not

M

comparable to corporate firms (Blume et al., 1998). The data sampling procedure is summarized

te

d

in Table 2.

Ac ce p



Further robustness checks of the results are tested to see if there is a negative association between credit risk rankings generated by structural models IJV extensions and their original counterparts and credit ratings. Firstly, credit ratings and survival probabilities can be converted into a comparable credit scores, followed by an explanation of how average credit scores can be used to test if IJV extensions of the structural bond credit risk models, generate lower survival probabilities than simply the conventional credit risk models. The correlation coefficient and 15

Issuer’s prefer debt type structures for IJV bonds because of their ease of structuring and similarity with conventional bonds. Moreover, they prefer to call them IJV bonds rather than debt bonds because IJV bonds are considered ideal from an Islamic perspective (Usmani, 2002; Ayub, 2007; AAOIFI, 2008) 20

Page 20 of 60

Kendall’s tau techniques are then used to measure the association between credit score results to test if there is a negative relationship between bond credit risk rankings generated by structural

ip t

models’ IJV extensions and their original conventional credit ratings and rankings.

cr

4.6.1 Credit Scores

The survival probability is a quantitative measure ranging from 0 to 1. As issuer’s have different

us

credit rating classification ranging from AAA, AA, A, BBB, BB, B, or lower (Standard and Poor’s, 2003), assigning of a quantitative “credit score” to both credit ratings and survival

an

probabilities would allow for a meaningful comparison (Finger et al., 2002). 16 Given their credit

(28)

(29)

Ac ce p

te

d

M

ratings and survival probability (PS), each issuer is scored from one to four as follows.

16

This is based on a Standard and Poor’s (2003) report which presents the default frequency (or survival frequency) for bonds in different rating categories. Default frequency = 1 – survival probability 21

Page 21 of 60

4.6.2 Averages Credit Scores The credit scores, using the above method, can be generated for the original structural models, their IJV extensions and issuer’s credit ratings. To test if the IJV extensions generate lower

ip t

survival probabilities than the conventional structural models alone, the average credit score can

(30)

an

us

cr

be calculated as follows.

where N is the number of firms in the sample. If the score generated by the IJV extensions are

M

lower than those of the original models and the issuer’s credit ratings, then this would support H1 in that the IJV extensions are lower than the original conventional bond credit risk models.

te

d

4.6.3 Correlation Coefficient and Kendall’s Tau The association between the credit scores as measured by their correlation coefficients and

Ac ce p

Kendall’s tau can test H2. These two approaches will now be discussed in turn.

A correlation coefficient measures the degree of linear dependence between two variables and so captures the extent to which both move together. For example, if the correlation between model X and model Y credit scores is around 1, firms assigned a high credit score by model X will have a high credit score from model Y. Alternatively with a negative correlation nearing -1, firms with a high credit score from model X will have a low credit score from model Y. A negative coefficient between the credit score of IJV extensions and the original models along with the credit ratings would support H2. This is because firms with the original model or credit ratings 22

Page 22 of 60

are assigned a low score by their IJV extension. The correlation between two variables X and Y can be calculated as follows:

is the covariance of variables X and Y, while

and

ip t

(31)

are their standard deviations,

cr

respectively. A problem, however, with using correlation coefficient is that outlier data and

us

extreme observations can greatly influence this measure. To correct for this problem another

an

measure of the association, Kendall’s tau, will be used (Finger et al., 2002).

Kendall’s tau measures the similarity between the rank ordering of two sets of variables (Finger

d

M

et al., 2002). For two random variables X and Y, the Kendall’s tau is defined as:

(32)

te

τ= P{(X2 − X1)(Y2 − Y1) ≥0} − P{(X2 − X1)(Y2 − Y1) <0}

Ac ce p

Where (X1, Y1) and (X2, Y2) are two observations generated from the joint distribution of X and Y. For our study, X is the original structural model credit score and Y is the IJV extension credit score. The first term in Equation 32 is the probability that original model and its IJV extension would rank a pair of firms in the same order. The second term can be interpreted as the probability of different rankings. A non-parametric estimation for Kendall’s tau17 is given by

17

The original version of Kendall’s tau is given by the following equation

A problem, however, with the above version of Kendall’s tau is that it reaches the extreme values of 1 and -1 only when all pairs are either concordant or discordant but with no tied ranks. In the presence of tied ranks the endpoint 23

Page 23 of 60

(Finger et al., 2002):

is the number of tied rank observations for X and

an

Here

us

cr

ip t

(33)

are the number of tied rank

observations for Y. Equation 33 calculates the excess number of concordant pairs of firms over

M

discordant pairs18. A concordant pair, which has the same relative ordering of X and Y, gets a

d

score of 1. A discordant pair, where the order ranking differs for X and Y is given a score of -

te

1.When there is neither concordance nor discordance, it gets a score of zero. A negative and statistically significant Kendall’s tau would be considered as supporting H2. The next section

5.

Ac ce p

presents the credit score results.

Results

This section reports on the survival probability simulation results for Merton (1974), first passage and CreditGrades models along with their IJV extension. These results enables H1 to be

values of 1 and -1 are not reached. A slightly modified version of Kendall’s tau as shown in Equation 33, therefore, is used in the paper (Kendall’s tau-b in EView’s manual). 18 This rescaling ensures that . Note that in the absence of ties, the modified version and the original version would yield similar results (Finger et al., 2002). 24

Page 24 of 60

tested to see if the IJV extensions of the structural models generate significantly lower survival probabilities than the original conventional models.

ip t

5.1 Simulation Results for Merton Model and its IJV Extenstion Survival probability results for the Merton (1974) model and its IJV extension are presented

cr

hereand were simulated using Excel. The results reflect changes in term to maturity followed by

us

results for stock volatility extensions as mentioned in Section 3.1.2.19

an

5.1.1 Term to Maturity (Merton Model)

M

The simulation results for the survival probability generated against term to maturity are presented in Figure 1, Panel A. It presents a standard case for a firm with low stock and for a firm with high stock volatility Debt (D) is set at 0.75 and equity

d

volatility

Ac ce p

te

(taken as debt/equity ratio of 40:60) at 1 as previously used by Finger et al. (2002).



The results for the Merton (1974) model shown in Figure 1 suggest that firms with low debt are likely to have almost perfect survival probability, irrespective of their term to maturity. The IJV extension results, however, indicate that if a firm with similar characteristics issues an IJV bond it will experience a relatively low survival probability. 20 This difference between the two models’ survival probabilites disappear as the bond’s term to maturity increases. The 19 20

The results presented here were also found to be robust to changes in other variables in the model. In the IJV case it is the probability of repayment of the principal. 25

Page 25 of 60

convergence is achieved at terms to maturity of around 10 years (Figure 1, Panel A2) 21 . In contrast, for a firm with high stock volatility, the convergence is only achieved at bond maturities in excess of 30 years. This discrepency could be attributed to IJV bonds having a

ip t

strong equity link. The stock volatility impact is greater in the short-term but smooths out with

cr

an increase in IJV bond maturity.

us

5.1.2 Stock Volatility (Merton Model)

The survival probability results generated against changes in stock volatility for a ten year bond,

an

holding terms to maturity constant, are shown in Figure 1, Panel A3. This indicates that the survival probability generated by the original Merton (1974) model remains unchanged with

M

increased stock volatility. In contrast, the IJV extension suggests that the survival probability for

d

IJV bonds decreases considerably with an increase in stock volatility. This discrepancy could be

te

attributed to IJV bonds having a stronger equity link. The results presented in Figure 1 Panel A, above support H1 for the Merton Model, suggesting that its IJV extension generates significantly

Ac ce p

lower survival probabilities than the original model. The next section analyses the results for the first passage model and its IJV extension.22 5.2 Simulations Results for First Passage Model and its IJV Extenstion The survival probability results for the first-passage model and its IJV extension generated against term to maturity and stock volatility are presented in this section.23

21

The results were also generated for a firm with high stock volatility. There, the convergence was achieved at bond maturities in excess of 30 years. This discrepency again relects IJV bonds having a strong equity link. The stock volatility impact was greater in the short-term but smooths out with an increase in IJV bond maturity. 22 The results presented here are robust to changes in stock volatility, debt ratio, equity ratio and a. 23 The results presented here are robust to changes in debt (D) and equity (S) variables in the model. 26

Page 26 of 60

5.2.1 Term to Maturity (First Passage Model) The simulation results depicted in Figure 1, Panel B1 show that the original first passage model

ip t

attributes high survival probability to firms with low debt, irrespective of the bond’s term to maturity. In contrast, their IJV extension assigns significantly lower firm’s survival probabilities.

cr

Unlike the Merton (1974) model, the difference between the survival probabilites of the original

us

first passage model and its extension remains considerable even for bonds with maturity in excess of 30 years. This implies that the effect of stock-volatility in the model has not smoothed

M

5.2.2 Stock Volatility (First Passage Model)

an

out overtime and so survival probabilities have not converged.

d

Figure 1, Panel B3 indicates that survival probability generated by the original first passage

te

model decreases only slightly at very high levels of stock volatility. The result for their IJV extension, however, suggests that survival probability decreases considerably with increases in

Ac ce p

stock volatility. Panel B3 also indicates that convergence in survival probabilities between the original model and its extension is only achieved at very low levels (around 0.1) of stock volatility. Moreover, with increasing stock volatility their survival probability diverges. The results presented here support H1 suggesting that there is a considerable difference in the survival probabilities of the original first passage model and its IJV extension.

27

Page 27 of 60

5.3 Simulations Results for CreditGrades Model and its IJV Extenstion This section present survival probability results for CreditGrades Model and its IJV extension

ip t

generated against term to maturity and stock volatility.24

5.3.1 Term to Maturity(CreditGrades Model)

cr

The results in Figure 1, Panel C1 show that survival probabilities generated by the original

us

CreditGrades model are higher than its IJV extension. These results, however, suggest that the misevaluation against the IJV extension is less severe than that of the Merton (1974) and first

an

passage models. The previous results in Section 5.1 and 5.2 also showed that a change in maturity had little or no impact on the original models’ survival probability for a low debt firm.

M

In contrast the Figures 1, Panel C1 above indicate that in the CreditGrades model, even for a low

d

debt firm, longer maturity is accompanied by lower survival probability.

te

The results for CreditGrades IJV extension also suggest that an increase in bond maturity leads

Ac ce p

to lower IJV bonds’ survival probability. This appears to be in slight contradiction to the Merton (1974) and first passage IJV extensions which suggested that higher maturity results in higher IJV survival probability. The rationale for this behaviour was that higher IJV bond maturities tend to smooth out the effect of the short-term stock volatility. A closer look at the Figure 1, Panel C1, however, suggests the slope of the IJV curve decreases with increasing levels of terms to maturity. The flattening of the slope for the IJV becomes considerable for maturities in excess of ten years. In contrast, the slope for original CreditGrades model’s survival probability remains

24

The results presented here are robust to changes in other variables in the model. 28

Page 28 of 60

relatively constant and decreases only slightly at higher maturities. This suggests that in CreditGrades IJV extension the effect of the stock volatility takes longer to smooth out.

ip t

5.3.2 Stock Volatility (CreditGrades Model)

Figure 1, Panel C3 shows that as stock volatility increase the survival probability generated by

cr

both the original CreditGrades model and its IJV extension decreases. The decrease, however, is

us

greater for the IJV extension. These results therefore support the H1 for the CreditGrades model because the IJV extension results in a lower survival probability than the conventional

an

CreditGrades model.

M

All the results discussed in Sections 5.1, 5.2 and 5.3 above show that the survival probability

d

generated by the IJV extension of the Merton (1974), first passage and CreditGrades models are

te

significantly lower than their original conventional counterparts. These results therefore fully support the H1. That is the structural models IJV extensions generate significantly lower survival

Ac ce p

probabilities than the original models. 5.4 Credit Score Results

The credit scores generated using credit ratings and the original structural models are presented in Table 3. The results for their IJV extensions are presented in Table 4.



29

Page 29 of 60

5.4.1 Average Credit Score Results The average credit scores results from Tables 3 and 4 are provided in Table 5. They show that

ip t

the average credit scores generated by the IJV extensions of the three structural models are significantly higher than their original counterparts. While Merton (1974) and the first passage

cr

IJV extension attributed the highest credit score of 4 to all the issuers, the misevaluation shown

us

by the CreditGrades IJV extension ranks them lower. However, its average credit score (3.42) compared with the original CreditGrades model (2.06) and credit ratings (1.83) is still

an

considerably high. So if a bond is rated by the CreditGrades IJV extension rather than the original model, its rating would be at least one grade less. Moreover, if rated by the Merton

M

(1974) and the first passage models IJV extensions, it declines by at least two grades. This supports H1 suggesting that the IJV extensions of the structural models attribute significantly

te

d

lower survival probability, credit ratings and credit score than the original models.

Ac ce p



5.4.2 Correlation Coefficient and Kendall’s Tau Results The correlation coefficients between the credit scores of the original models and credit ratings are provided in Table 6. Table 7, moreover, shows the correlations for the CreditGrades IJV extension.25 The Kendall’s tau results are displayed in Table 8 and 9.

25

Merton and first passage models IJV extensions assigned all the issuers the maximum score of 4. As there is no variation in their credit score series, their correlation coefficients and Kendall’s tau with other models could not be calculated. 30

Page 30 of 60



ip t



cr

Table 6 shows that all original structural models have positive correlation coefficients with the

us

credit ratings. The Kendall’s taus (Table 8) are also positive. This implies a positive association between the survival probabilities of the original models and credit ratings and so corroborates

an

the assertion that credit rating agencies are currently using conventional credit risk models when rating Islamic bonds.26 The CreditGrades model shows the highest correlation coefficient and the

te

d

Merton (1974) and first passage models.

M

Kendall’s tau with credit ratings. This suggests superiority of the CreditGrades model over the

The CreditGrades IJV extension’s correlation coefficient (shown in Table 7) and Kendall’s tau

Ac ce p

(shown in Table 9) with both credit ratings and the original CreditGrades model are positive, statistically significant and has a reasonably high magnitude. 27 This shows that they have a positive association. It further implies that bond issuers assigned a high credit score by the CreditGrades model will also receive a high score by its IJV extensions. These results, therefore, do not support H2.

26

While AAOIFI has declared that most of the IJV bonds are not equity based, there is evidence that some Islamic bonds were more equity in nature. For example, Sitara Chemicals issued sukuk in Pakistan. The number of IJV bonds is small. The strong bias against equity type structures could have deterred their issuance (AAOIFI, 2008). 27 The correlation coefficient and Kendall-tau with the IJV extensions of Merton (1974) and the first passage cannot be calculated as the variance of their credit score series is zero. 31

Page 31 of 60

6. Conclusion This paper examined the credit risk nature of a specific sukuk type, the IJV bonds. It considered its unique position in respect to risk and returns and noted that IJV bonds do not promise the

ip t

safekeeping of principal and instead offer a return in line with issuer’s performance. A default in an IJV context, therefore, is quite different from conventional debt and its return on principal is a

us

cr

function of IJV bond’s market value.

Given these differences, this paper analysed whether the conventional credit risk models can

an

effectively assess the risk in IJV bonds. We attempted to enhance the conventional models (Merton, first passage, and CreditGrades) with extensions designed to reflect IJV characteristics.

M

The methodology for comparing credit rating and survival probabilities using credit scores was

d

then discussed, followed by the averages, correlation coefficients and Kendall’s tau methods to

te

analyze credit scores. Both the original models and the extensions were then used to run simulations. The results supported our H1 that IJV extensions of structural bond credit risk

Ac ce p

models generate significantly lower survival probabilities than these original conventional credit risk models. The testing of Malaysian bond data was then introduced to check the robustness of previous results. The methodology for comparing credit rating and survival probabilities using credit scores was then discussed, followed by the averages, correlation coefficients and Kendall’s tau methods to analyze credit scores. Finally, the testing of real Malaysian data results clearly supported the previous H1 conclusion that the IJV extensions generate significantly lower

32

Page 32 of 60

survival probabilities than the original models. They did not however, support H2, implying that these models can be used as an additional tool to rank order IJV bonds.28

ip t

The positive testing of H1 implies that when IJV bonds compete against debt bonds for credit ratings, these structural models and their extensions will misevaluate them as they will attribute

cr

lower survival probabilities to IJV structures, they will then assign lower credit ratings. The

us

negative testing of H2, however, suggests a positive association between the CreditGrades model and its IJV extension. A negative association between them would have reflected a misevaluation

an

suggesting that firms ranked higher, in terms of their riskiness, by CreditGrades are ranked lower by their IJV extensions. A positive association, however, reflects that in terms of rankings both

M

the original models and IJV extensions agree on the firm’s relative riskiness.29 Therefore, if IJV

d

bonds are considered as distinct from debt bonds, than structural IJV extension models can be

te

used as an adjunct tool for rating them. However, this analysis would also need to consider the equity type nature of IJV bonds, including the entitlement to high positive returns, in case the

Ac ce p

issuer performs well.

A major problem we identified is that conventional credit risk models focus on the issuer’s ability to return the principal but ignore any additional positive returns that IJV bonds may generate. This mispricing cannot be addressed by merely extending the conventional models to capture some of the IJVs unique features. Our conclusion is that IJV bonds would be better 28

Note that IJV bonds can be used for both project finance and also full venture finance but would not change the outcome of the model or the results. The model included a variable that captures how credit risk might change if IJV bonds are used to finance particular projects.

29

Note that this ability to differentiate is only in relative and not absolute terms as it has already been shown in testing H2 that in absolute terms the IJV extensions assign issuers lower survival probability and higher risk. 33

Page 33 of 60

evaluated like equity rather than a debt investment. Hence, tools used to measure equity risk should be applied to IJV bonds.

ip t

The implications of our findings should be viewed rather more widely than simply how to best evaluate IJV bonds. They point to what appears to some unfortunate short comings in Islamic

cr

finance or at least in the way it has developed. There is nothing wrong in fact probably has

us

considerable advantages, in starting with conventional finance in the first instance when designing new Islamic products. The problem is not in fully understanding the implications that

an

any specific Islamic terms and conditions might have on their risk and return parameters. If this was done and that information widely distributed then Islamic specific, risk assessment models

M

would have already been widely used Fortunately, many of these matters are now being

te

own Basel III equivalent.

d

addressed as part of the Islamic financial industry and its regulators attempts to implement their

Ac ce p

In terms of future research, two suggestions are offered. The first relates to the problem found with the conventional credit models and their assumptions. The IJV extensions afford here are just the first step in the process. Researchers have a wealth of opportunities to make a similar assessment to the intrinsic nature of other Islamic instruments and modify conventional models to reflect their constraints and outcomes. This approach is not limited simply to Islamic capital market instruments but should include Islamic bank loans and deposit instruments, too. The second aspect is to examine the position of the IJV more broadly and assess not only the reasons for its relatively lower usage but also how it might be used more effectively both by Islamic banks

and

in

the

capital

markets. 34

Page 34 of 60

REFERENCES: June

13,

2009,

from

ip t

AAOIFI., 2008. Sukuk Resolution. Retrieved http://www.aaoifi.com/aaoifi_sb_sukuk_Feb2008_Eng.pdf

cr

Acharya, V., Naqvi, H., 2012. The seeds of a crisis: A theory of bank liquidity and risk taking over the business cycle. Journal of Financial Economics 106, 349–366.

us

Ammann, M., 2001. Credit Risk Valuation: Methods, Models, and Applications (2nd ed.). New York, NY: Springer. Ayub, M., 2007. Understanding Islamic Finance. Chichester, UK: John Wiley and Sons.

an

Benartzi, S., Thaler, R., 1995. Myopic loss aversion and the equity premium puzzle. Quarterly Journal of Economics 110, 73-92.

M

Black, F., Cox, J.C., 1976. Valuing corporate securities: some effects of bond indenture provisions. Journal of Finance 31, 351-367.

d

Black, F., Scholes, M., 1973. The pricing of options and corporate liabilities. Journal of Political Economy 81, 637-654.

te

Blume, M.E., Lim, F., MacKinlay, A.C., 1998. The declining credit quality of us corporate debt: Myth or reality? Journal of Finance 53, 1389-1413.

Ac ce p

Bolton, P., Freixas, X., Shapiro, J., 2012. The credit ratings game. Journal of Finance 67, 85– 111. Brennan, M.J., Schwartz, E.S., 1980. Conditional predictions of bond prices and returns. The Journal of Finance 35, 405-417. Briys, E., de Varenne, F., 1997. Valuing risky fixed rate debt: An extension. Journal of Financial and Quantitative Analysis 32, 239-248. Brunnermeier, M., 2009. Deciphering the liquidity and credit crunch 2007–08. Journal of Economic Perspectives, 23, 77–100. Chong, B.S., Liu, M.H., 2009, Islamic banking: Interest-free or interest-based? Pacific-Basin Finance Journal 17, 125–144. Duffie, D., Singleton, K.J., 1999. Modelling term structures of defaultable bonds. Review of Financial Studies 12, 687-720.

35

Page 35 of 60

Duffie, D., Singleton, K.J., 2003. Credit Risk: Pricing, Management and Measurement. Princeton, New Jersey: Princeton University Press. Ernst & Young (EY), 2014. World Islamic Banking Competitiveness Report 2013–14: The transition begins. Ernst & Young.

ip t

Finger, C.C., Finkelstein, V., Lardy, J.P., Pan, G., Ta, T., Tierney, J., 2002. CreditGrades Technical Document. New York, NY: RiskMetrics Group.

cr

Howladar, K., 2006. Shariah and Sukuk: A Moody’s Primer. (International Structured Finance Special Comment). New York, NY: Moody’s Investors Service.

us

Jarrow, R.A., Turnbull, S.M., 1995. Pricing derivatives on financial securities subject to credit risk. Journal of Finance 50, 53-85.

an

Jarrow, R.A., Lando, D., Turnbull, S. M., 1997. A Markov model for the term structure of credit risk spreads. Review of Financial Studies10, 1-42.

M

Kahnemen, D., Tversky, A., 1979, Prospect Theory: An analysis of decision under risk. Econometrica 47, 263-292. Khan, F., 2010. How ‘Islamic’ is Islamic banking? Journal of Economic Behavior & Organization 76, 805–820.

te

d

Lando, D., 2004. Credit Risk Modelling: Theory and Applications. Princeton, New Jersey: Princeton University Press.

Ac ce p

Longstaff, F.A., Schwartz, E.S., 1995. A simple approach to valuing risky fixed and floating rate debt. Journal of Finance 50, 789-819. Meissner, G., 2005. Credit Derivatives: Application, Pricing, and Risk Management. Malden, MA: Blackwell Publishing. Merton, R.C., 1974. On the pricing of corporate debt: The risk structure of interest rates. Journal of Finance 29, 449-470. Quantitative Micro Software, 2010. EViews 6 User's Guide. Irvine, CA: Author. Standard and Poor's, 2003. Standard and Poor's Corporate Ratings Criteria. New York, NY: Standard and Poor’s. Usmani, T.M., 2002. An Introduction to Islamic Finance. Hague, Netherlands: Kluwer Law International. Usmani, T.M., 2007. Sukuk and their contemporary applications, The Shariah Council.

36

Page 36 of 60

Appendix A: Merton Model IJV Extension

(A.1)

ip t

(A.2)

cr

(A.3)

(A.5)

(A.6)

(A.7)

Ac ce p

te

d

M

an

us

(A.4)

(A.8)

(A.9)

(A.10)

37

Page 37 of 60

First Passage IJV Extension

ip t

(A.11)

cr

(A12)

an

us

(A.13)

(A.14)

M

For simplicity assume that

d

(A.15)

Ac ce p

te

(A.16)

(A.17)

(A.18) (A.19) (A.20)

(A.21)

38

Page 38 of 60

cr

ip t

(A.22)

Ac ce p

te

d

M

an

us

CreditGrades IJV Extension

(A.23) (A.24) (A.25) (A.26) (A.27)

(A.28)

(A.29)

(A.30)

39

Page 39 of 60

Table 1: Islamic Bond Structures

SARA bonds

!! !!

Debt bonds

!!

!!

Represent ownership in a particular project rather than the whole venture. Have a maturity date. Islamic bonds represent ownership of underlying asset. Use the concept of lease financing. They have to be structured based on buying and reselling of the underlying asset so that a debt is created. The Islamic bank first buys a real asset and then immediately sells it back to the client at a higher price creating a debt which the client periodically repays.

!!

Musharkah

!!

Ijarah

!!

Murabaha

ip t

!!

Islamic Type of Contract

cr

Additional Islamic Constraints

us

Islamic Instrument IJV bonds

Ac ce p

te

d

M

an

Source: Adapted from Usmani (2002) and AAOIFI (2008).

40

Page 40 of 60

Table 2: Sampling Procedure and Sample Size This table outlines the sample selection criteria and total number of sample bond issuers considered. The data is for Malaysian Islamic bond issuers from the period 2002 to 2010 taken from IFIS data base. Description

No

Total corporate Islamic bond issuers obtained from the IFIS data base

81

ip t

Exclude issuers with missing data in the IFIS

244

Exclude non-Malaysian issuers.

33

30

47

cr

Exclude issuers with missing daily data in the Datastream Exclude issuers with missing in Compustat

52

Ac ce p

te

d

M

an

us

Final sample

31

30

A number of issuers were unlisted on KLSE. 41

Page 41 of 60

Table 3: Credit Scores from Original Models and Credit Ratings This table presents credit scores generated using credit ratings, Merton (1974), first passage and CreditGrades model. The data used is for Malaysian Islamic bond issuers from the period 2002 to 2010 taken from IFIS data base.

d

KINSTEEL

Debt

2

CreditGrades

1

1

1

3 1 1 1 1 1 1 3 1 3 2 1 4 1 2 4 4

ip t

2 1 1 1 1 1 1 3 1 1 2 1 3 1 2 1 3

First Passage 3 2 2 2 2 1 1 4 1 2 2 1 4 1 3 3 4

cr

Merton

Debt

2

4

4

4

Debt

1

1

1

1

LAFARGE MALAYAN CEMENT

Debt

2

2

3

2

LEADER UNIVERSAL

Debt

2

3

4

4

Ac ce p

KWANTAS CORP

te

KNM GROUP

SARA Debt Debt Debt IJV Debt Debt Debt Debt Debt IJV Debt Debt IJV Debt Debt SARA

M

ALAM MARITIM ATLAN BINA DARULAMAN BOON KOON GROUP CHEMICAL COMPANY DELLOYD VENTURES DRB-HICOM EMAS KIARA IND ESSO MALAYSIA EVERMASTER GROUP GAMUDA GLOMAC GOODWAY INTEGRATED HONG LEONG INDUSTRIES HUBLINE HYTEX INTEGRATED INGRESS CORPORATION

Credit Ratings 1 2 1 2 1 1 1 2 1 3 1 2 2 1 1 3 4

us

Type

an

Company

LINGKARAN TRANS KOTA

IJV

1

1

1

1

MALAYSIAN AE MODELS

Debt

2

3

4

4

MALAYSIAN MER.MAR.

Debt

3

1

2

2

MATAHARI PUTRA PRIMA

SARA

2

2

4

3

MAXTRAL INDUSTRY

Debt

2

4

4

4

MIECO CHIPBOARD

Debt

1

1

1

1

MINETECH RESRCS

Debt

2

2

3

3

MISC

Debt

1

1

1

1

MUHIBBAH ENGINEERING

IJV

2

1

1

1

MULPHA INTERNATIONAL

Debt

1

1

1

1

NAM FATT CORP

Debt

3

1

1

3

NESTLE (MALAYSIA)

Debt

1

1

1

1

OILCORP

Debt

4

1

1

4

PHARMANIAGA

Debt

1

1

1

1

PLUS EXPRESSWAYS

IJV

1

1

1

1 42

Page 42 of 60

43

Page 43 of 60

d

te

Ac ce p us

an

M

cr

ip t

Table 3: Credit Scores from Original Models and Credit Ratings Continued Merton

First Passage

Debt Debt Debt Debt SARA Debt Debt Debt Debt Debt Debt Debt IJV Debt Debt

1 2 2 2 1 1 2 2 1 1 4 1 1 1 2

1 4 3 1 1 3 1 3 1 1 2 1 1 1 3

2 4 4 1 1 4 1 4 1 1 3 2 1 1 4

CreditGrades

us

cr

ip t

Credit Ratings

1 4 4 1 1 4 1 3 1 1 3 2 1 1 4

Ac ce p

te

d

M

POH KONG PRICEWORTH WOOD PRDS. PRINSIPTEK CORP RANHILL SAPURACREST PETROL SCOMI GROUP SUNRISE SYMPHONY HOUSE TENAGA NASIONAL TOP GLOVE CORP TRACOMA TRADEWINDS CORP UMW HOLDINGS WEIDA (M) ZECON

Type

an

Company

44

Page 44 of 60

Table 4: Credit Scores from IJV Extensions

Debt

4

4

BINA DARULAMAN

Debt

4

4

BOON KOON GROUP

Debt

4

4

CHEMICAL COMPANY

IJV

4

DELLOYD VENTURES

Debt

4

DRB-HICOM

Debt

4

EMAS KIARA IND

Debt

ESSO MALAYSIA

Debt

EVERMASTER GROUP

4 4 4

cr

ATLAN

ip t

This table presents credit scores generated using IJV extensions of Merton (1974), first passage and CreditGrades model. The data used is for 52 Malaysian Islamic bond issuers from the period 2002 to 2010 taken from IFIS data base. Company Type Merton First CreditGrades IJV Passage IJV IJV ALAM MARITIM SARA 1 4 4

4

4

4

4

1

4

4

4

4

4

2

Debt

4

4

4

GAMUDA

IJV

4

4

4

GLOMAC

Debt

4

4

4

GOODWAY INTEGRATED

Debt

4

4

4

IJV

4

4

Debt

4

4

4

an

HUBLINE

KNM GROUP

4

4

4

4

4

4

Debt

4

4

1

Debt

4

4

4

Debt

4

4

1

LAFARGE MALAYAN CEMENT

Debt

4

4

4

LEADER UNIVERSAL

Debt

4

4

4

LINGKARAN TRANS KOTA

IJV

4

4

2

MALAYSIAN AE MODELS

Debt

4

4

4

MALAYSIAN MALAYSIAN MERCHANT MARINE. MATAHARI PUTRA PRIMA

Debt

4

4

4

SARA

4

4

4

MAXTRAL INDUSTRY

Debt

4

4

4

MIECO CHIPBOARD

Debt

4

4

3

MINETECH RESRCS

Debt

4

4

4

MISC

Debt

4

4

1

MUHIBBAH ENGINEERING

IJV

4

4

3

MULPHA INTERNATIONAL

Debt

4

4

4

NAM FATT CORP

Debt

4

4

4

NESTLE (MALAYSIA)

Debt

4

4

1

Ac ce p

KWANTAS CORP

Debt

SARA

te

KINSTEEL

4

d

HYTEX INTEGRATED

M

HONG LEONG INDUSTRIES

INGRESS CORPORATION

us

4

45

Page 45 of 60

Merton IJV

First Passage IJV

CreditGrades IJV

OILCORP

Debt

4

4

4

PHARMANIAGA

Debt

4

4

PLUS EXPRESSWAYS

IJV

4

4

POH KONG HOLDINGS

Debt

4

4

PRICEWORTH WOOD PRDS.

Debt

4

4

4

PRINSIPTEK CORP

Debt

4

4

4

RANHILL

Debt

4

4

4

4

1

4

4

4

4

4

4

4

4

4

4

4

4

1

Debt

4

4

4

Debt

4

4

4

4

Debt

4

SUNRISE

Debt

4

SYMPHONY HOUSE

Debt

TENAGA NASIONAL

Debt

TOP GLOVE CORP

Debt

TRACOMA HOLDINGS

IJV

4

4

1

4

4

4

Debt

4

4

4

Ac ce p

ZECON

4

Debt

te

UMW HOLDINGS

2

d

TRADEWINDS CORP WEIDA (M)

an

SARA

SCOMI GROUP

M

SAPURACREST PETROL

1

cr

Company

ip t

Type

us

Table 4: Credit Scores from IJV Extensions Continued

46

Page 46 of 60

Table 5: Average Credit Scores

Ac ce p

te

d

M

an

us

cr

ip t

This table presents the average credit scores results given in Tables 3 and 4. The data used is for 52 Malaysian Islamic bond issuers from the period 2002 to 2010 taken from IFIS data base. Original IJV Extension 1.6538 4 Merton 2.1538 4 First Passage 2.0962 3.32692 CreditGrades 1.7115 Credit Ratings

47

Page 47 of 60

ip t

Table 6: Correlation Coefficients between Original Structural Model and Credit Ratings

Merton First Passage CreditGrades

0.2582 0.3548 0.5897

1 0.8978 0.8080

1 0.8480

us

1

1

Ac ce p

te

d

M

an

Credit Ratings

cr

This table presents the correlation coefficients between credit scores results given in Table 4. The data used is for 52 Malaysian Islamic bond issuers from the period 2002 to 2010 taken from IFIS data base. Credit Ratings Merton First Passage CreditGrades

48

Page 48 of 60

Table 7: CreditGrades, IJV Extension and Credit Ratings Correlation

Ac ce p

te

d

M

an

us

cr

ip t

This table presents the correlation coefficients between credit scores results generated using credit ratings, CreditGrades model and its IJV extensions. The data used is for 52 Malaysian Islamic bond issuers from the period 2002 to 2010 taken from IFIS data base. CreditGrades IJV 0.4084 Credit Ratings 0.4997 CreditGrades

49

Page 49 of 60

Table 8: Kendall’s Tau between Original Structural Model and Credit Ratings

0.5373

1.0000

cr

CreditGrades

ip t

This table presents the Kendall’s tau between credit scores results given in Table 3. The data used is for 52 Malaysian Islamic bond issuers from the period 2002 to 2010 taken from IFIS data base. The number in the parenthesis provides the probability for testing S = 0, implying that there is no association between the different credit scores. Credit Ratings CreditGrades First Passage Merton 1.0000 Credit Ratings

(0.0000) 0.7793

(0.0022)

(0.0000)

0.3149

0.7771

(0.0131)

(0.0000)

1.0000 0.8297

1.0000

(0.0000)

Ac ce p

te

d

M

an

Merton

0.3793

us

First Passage

50

Page 50 of 60

Table 9: Kendall’s Tau between CreditGrades, IJV Extension and Credit Ratings

cr

ip t

This table presents the Kendall’s tau between credit scores results generated using credit ratings, CreditGrades model and its IJV extensions. The data used is for 52 Malaysian Islamic bond issuers from the period 2002 to 2010 taken from IFIS data base. The number in the parenthesis provides the probability for testing S = 0, implying that there is no association between the different credit scores CreditGrades IJV 0.424849 Credit Ratings

us

(0.001)

0.491162

CreditGrades

Ac ce p

te

d

M

an

(0.0001)

51

Page 51 of 60

cr us

Figure 1: Simulation Results

2

3 Against Stock Volatility

ed

M

an

1 Panel A: Merton Model: Survival Probablity Against Term to Maturity High volatility firm Low volatility firm

  Against Stock Volatility  

Ac

ce

pt

  Panel B: First Passage Model: Survival Probability Against Term to Maturity Low volatility firm High volatility firm

 

  Panel C: CreditGrades Model: Survival Probability Against Term to Maturity Low volatility firm High volatility firm

 

 

  Against Stock Volatility  

 

 

 

Page 52 of 60

cr

Merton Model: Survival Probability Against Term to Maturity (High Volatility Firm)

Figure 3:

Merton Model: Survival Probability Against Stock Volatility

Ac

ce

pt

ed

M

an

us

Figure 2:

Page 53 of 60

cr us an M ed pt ce Ac

Figure 4:

First Passage Model: Survival Probability Against Term to Maturity (Low Volatility Firm)

Page 54 of 60

cr us an M ed pt ce

 

First Passage Model: Survival Probability Against Term to Maturity (High Volatility Firm)

Ac

Figure 5:

Page 55 of 60

cr us an M ed pt Ac

ce



Figure 6:

 

First Passage Model: Survival Probability Against Stock Volatility

Page 56 of 60

cr us an M ed pt ce Ac

Figure 7:



 

CreditGrades Model: Survival Probability Against Term to Maturity (Low Volatility Firm)

Page 57 of 60

cr us an M ed pt Ac

ce



Figure 8:

 

CreditGrades Model: Survival Probability Against Term to Maturity (High Volatility Firm)

Page 58 of 60

cr us an M ed pt ce Ac

Figure 9:



 

CreditGrades Model: Survival Probability Against Stock Volatility

Page 59 of 60

cr us an M ed pt  

Ac

ce



 

Page 60 of 60