This action might not be possible to undo. Are you sure you want to continue?
Bond Implied CDS Spread and CDSBond Basis
Richard Zhou
†‡
August 15, 2008
Abstract
We derive a simple formula for calculating the CDS spread implied by the bond market
price. Using noarbitrage argument, the formula expresses the bond implied CDS spread
as the sum of bond price, bond coupon and Libor zero curve weighted by risky annuities.
We show that the bond implied CDS spread is consistent with the standard CDS pricing
model if the survival probabilities and recovery are consistent with the bond price.
1. Introduction
A CDS contract is an OTC transaction between two parties in which the protection buyer
pays a stream of coupon payment to the protection seller until the earlier of maturity or
entity default in exchange for a default contingent payment. The common default
settlement is the physical settlement where the protection buyer delivers a bond from a
pool of eligible bonds to the protection seller in exchange for par. CDS contracts can also
be cash settled where the protection buyer receives from the protection seller the cash
amount of par less recovery.
With the physical settlement, the CDS protection buyer holds a delivery option where he
can choose any bond from a pool of bonds to deliver into the CDS contract. Empirical
evidence shows that bonds of the same entity do not necessarily have the same market
value following default [1]. As a result, the standard CDS pricing with a flat recovery rate
cannot properly price in the value of the delivery option embedded in the CDS contracts.
Given the issuer default probability, the bond price is determined by the recovery and
other fundamental and market technical factors such as supply and demand, and liquidity.
From modeling perspective, it is difficult to separate recovery, default probability, and
other market fundamental and technical factors since they are intertwined. The recovery
swap prices can be used as the expected recovery rate but the market has not yet fully
developed. Even if the recovery rate can be determined independently, the default
probabilities calibrated to the market CDS spreads or bond prices are still contaminated
by other factors such as supply and demand, funding cost, bond trading away from par
(see [7] for a detailed exposition on factors impacting CDS and cash basis).
Empirical studies show that the markets apears to price CDS based on Libor curve rather
than the treasury curve [2]. CDS discounting should be based on Libor. Since the Libor
†
Risk Management, The Depository Trust & Clearing Corporation, Email: rzhou50@gmail.com.
‡
The opinions of this article are those of the author and do not reflect in any way the views or business of
his employer. All errors are author’s own.
2
are the borrowing rates between banks of AA rating, the Libor curve is implicitly an AA
rated yield curve. As a result, we use Libor as the risk free interest rate.
An asset swap (ASW) is a package transaction between two parties in which the ASW
buyer purchases a bond from the other party and simultaneously enters into an interest
rate swap transaction, usually with the same counterparty, to exchange the coupon on the
bond for Libor plus a spread. The spread is called the asset swap spread. A common asset
swap is the par asset swap where the buyer pays par at the inception of the deal. Unlike
CDS, ASW continues following bond default.
CDSBond basis is the difference between the CDS spread and the ASW spread on the
same bond. It is a general indicator of relative value of CDS versus the cash bond. For
example, when the CDS spread is higher than the ASW spread, i.e. the basis is positive,
the CDS is generally considered to be more attractive than the bond. The reverse is true if
the basis is negative.
Bond implied CDS spreads have been previously investigated. Davies and Pugachevsky
proposed an approximation method for calculating the bond implied CDS spread based
on the Z spread adjusted by the bond’s market price, duration, convexity and recovery
rate [3]. In a series of paper, Berd et al proposed to use the survivalbased modeling as a
consistent measure for creditrisky bond pricing and risk management [4]. In their model,
the survival probability term structure of an issuer is estimated by regressing model
prices against market prices across all bonds of that issuer under a constant recovery rate.
The resulting survival probability term structure is not necessarily the same as that
implied by the CDS market. The idiosyncratic fitting error is accounted for by the OAS
tofit which is essentially a measure in spread of aggregate effect of market factors on the
bond.
In this paper, we describe a simple model for estimating the bond implied CDS spread.
The model concept is not new as it is based on survival probability. But it is cast in such
an explicit way that it is easy to analyze the effects on the CDS spread of bond coupon,
recovery rate, bond price, and interest rate and credit curves. These effects have been
discussed in literature, but an explicit formula that combines all these factors seems
needed and this paper provide a such formula.
The paper is organized as follows. Section 2 describes the underlying theory and main
formulas. Section 3 gives some numerical examples, and section 4 concludes the paper.
The detailed model derivation is given in Appendices A and B.
2. CDS and Asset Swap Spreads
In this section, we describe a model for calculating the bond implied CDS spread and the
CDSBond basis. We define the bond implied CDS spread as the spread that equates the
bond market price with the bond fair theoretical price. We demonstrate the effects of
Libor curve, bond coupon and the price deviation from par on the credit spread.
3
2.1 Asset Swap Spread
In a par asset swap transaction, the investor buys a package consisting of a cash bond and
a payer interest swap where the investor swaps the bond coupon for Libor plus a (ASW)
spread. The price of the asset swap package is par (hence the term par asset swap) which
the investor pays at the deal inception. This means that the bond dirty price and the initial
swap value must sum to par. For a discount bond, the initial swap value is positive to the
investor. For a premium bond, it is negative. The market practice is asset swap does not
knock out when the underlying bond defaults. As a result, the investor bears some interest
rate risk in that he needs to seek new funding to pay the fixed rate should the bond
default.
Let
0
T be today, and the bond coupon payment dates be N k T
k
,..., 1 , = . Furthermore, we
assume that the ASW payment dates coincide with the bond coupon payment dates.
The spread of a par asset swap is given by
A
D
L C
A
D M
S
ASW
+ − =
− −
=
) 1 (
(1)
where M is the bond riskfree price, 1D is the bond’s dirty price, C is the bond coupon,
1 − k
L is
the −
−1 k
T maturity forward Libor rate paid at
k
T , and A is the riskfree annuity given by
( )
∑
=
−
∆ =
N
k
k k
T DF T A
1
1
(2)
where ( )
k
T DF is the discount factor for maturity
k
T ,
1 1 − −
− = ∆
k k k
T T T , and the riskfree
average Libor weighted by the discount factors is
( )
∑
=
− −
∆ =
N
k
k k k
T DF T L
A
L
1
1 1
1
(3)
The first form of ASW spread in formula (1) is well known. It states that the ASW spread is the
difference between the riskless price and the market price amortized over the swap life. Since the
bond riskfree price is always greater than the market price, ASW spread is always
positive. Note that the spread would have to be negative if the bond market price is
greater than the bond riskfree price.
The second form of formula (1) is less familiar, but more intuitive. It shows that the ASW spread
is composed of three components: the bond coupon, the average Libor over the swap life and the
difference of bond dirty price and par scaled by the riskless annuity. The last component can also
be interpreted as the bond discount amount amortized over the swap life. This discount amount is
paid upfront by the investor and recouped over the life of the swap as part of the swap spread.
4
Remarks:
1) If the bond is trading at par (D = 0), the ASW spread is the difference between the
bond coupon and the average Libor rate over the ASW term.
2) The ASW spread increases as the bond price decreases (D increases). Hence, the
more deeply discounted is the bond, the higher is the ASW spread.
3) We will show in the next section that the bond implied spreads – given by formula (7)
– has characteristics similar to the ASW spread. The bond implied spread contains
three terms with interpretations similar to those of terms in formula (1). And the bond
implied spread increases with decreasing bond price (increasing D) but at a faster rate
than ASW spread does, resulting in an increasing basis.
2.2 Bond Implied CDS Spread
Suppose an investor executes a socalled negative basis trade in which he buys the cash
bond on the ASW basis and buys CDS protection. The market price of the bond is 1D
where D represents bond’s discount relative to par. Assuming a recovery rate R, the
investor needs to buy 1D/(1R) notional CDS protection in order to make the combined
position default neutral. Given a nonzero recovery, this amount is less (more) than 1D
for a discount (premium) bond.
Let us assume that the investor borrows 1D to purchase a bond with fixed coupon C. He
then buys 1D/(1R) notional CDS protection on the bond. Denoting the recovery rate by
R, the bond implied CDS spread is given by (See Appendix A for derivation and
definitions of the terms in equations (4), (5) and (6)).
( )
)
`
¹
¹
´
¦
(
¸
(
¸
−
×
− + − =
D
Loss R
PV
D
L
PV
PV
C
W
S
Risky
CDS
1
1
01 01
01 1
(4)
where ( )
D
D
R
R
D
R
D
W
− −
− = − 
¹

\

−
− =
1 1
1 1 /
1
1 under the constraint
( ) ( )
N N
T DF T P Loss R PV C D × > + × + × = − τ 01 1 (5)
Equation (4) shows that the bond implied CDS spread is the sum of contributions from
the bond coupon, the Libor curve, the difference between the bond (dirty) price and par,
and the recovery rate augmented by the risky annuities and W which is the ratio of the
CDS notional amount to the bond price.
We will show in Appendix B that the bond implied CDS spread calculated using equation
(4) automatically satisfies the standard CDS pricing equation
5
Loss
PV
R
S
CDS
×
−
=
01
1
(6)
This implies that, if and only if the recovery and default probability are consistent with
the bond price, is the model consistent with the standard CDS pricing model (6). The
reverse does not hold. Given a CDS spread and a recovery rate R, the default probability
implied by equation (6) is generally inconsistent with equation (5).
Substituting equation (6) into equation (4) and solve for
CDS
S , we find a simplified form
01 01
01
PV
D
L
PV
PV
C S
Risky
CDS
+ − × = (7)
It is clear from formula (7) that for a par bond (D = 0) and a flat Libor curve, we have
L
PV
PV
C S
CDS
− =
01
01
which is slightly less than the CL implied by the risky par floater
replication model. The difference is due to that in our model, the bond accrued coupon is
not paid but the accrued CDS premium and Libor interest are paid upon default, while
there is no such default payment discrepancy in the par floater replication.
Given the bond’s market price 1D, the model framework of (57) can be used in several
ways:
1) Single bond: In this situation, we assume a recovery rate R, and calculate the constant
bond implied hazard rate using equation (5). The bond implied CDS spread is then
calculated using equation (7). The bond implied hazard rate is not necessarily
consistent with the hazard rate implied by the market CDS quote of the same issuer.
The bond implied hazard rates are determined by the fundamental and technical
factors in the cash market while the CDS hazard rates are determined by the CDS
market fundamental and technical factors. However, if we are concerned only with
the CDS spread, equation (7) is all that matters. Therefore, equation (7) can be used to
compare the bond implied CDS spread to the market CDS quote.
2) Multiple bonds of differing maturities: Assuming a constant recovery, we bootstrap to
obtain a term structure of the bond implied hazard rate consistent with the given bond
prices. The bond implied CDS spread for a maturity can then be calculated from (7).
3) Term structures of market CDS spread and bond price: We bootstrap to calculate the
hazard rate term structure and the bond implied recovery rate by simultaneously
solving equations (5) and (6). The resulting hazard rate and recovery rate term
structures are consistent with the bond market prices and the market CDS spreads of
the issuer. However, the recovery rates are influenced by the cash market factors and
are not necessarily the expected percentage recovery amount of par.
6
Remarks:
1) Interestingly, the CDS spread formula (7) can also be directly obtained if the CDS
notional is 1D. 1D/(1R) CDS notional is default neutral. 1D notional would result
in a small default payoff of DR which is offset by the larger carry (1D)*S.
2) The form of equation (7) is useful as it explicitly expresses the CDS spread in terms
of bond coupon, interest rate curve, and bond price augmented by the risky annuities,
allowing easy analysis of individual factors.
3) It is important to note that in equations (47) we have not imposed any restriction on
the shape of credit and interest rate curves.
2.3 CDS–Bond Basis
The CDS–Bond basis is defined as the difference between the bond implied CDS spread
(7) and the par asset swap spread (1)
( )


¹

\

− + 
¹

\

− − − =
A
PV
PV
D
PV
PV
C L L Basis
Risky
01
1
01 01
01
1 (8)
The CDSBond basis consists of three terms:
1) The first term is the difference between the average riskfree Libor rate and the
average risky Libor rate. This term increases with increasing bond discount D, or
increasing default risk. It is positive for an upward sloping interest rate curve, and
negative for a downward sloping curve. It is zero when the forward curve is flat.
Therefore, the first term can be interpreted as the impact of the yield curve slope to
the basis. Since the normal yield curve shape is upward sloping, the interest rate curve
effect on the basis is generally positive (see Table 1).
2) The second term is due to the payment mismatch between the bond coupon and CDS
premium and borrowing cost upon default. While the investor does not receive the
bond’s accrued interest in the event of default, he still needs to pay the accrued CDS
premium and interest on the loan. This term always contributes negatively to the
CDSBond basis.
3) The third term represents the effect of bond’s market price on the CDSBond basis. It
is positive for discount bond and negative for premium bond. It explains why, rough
speaking, CDSBond basis is positive for discount bond and negative for premium
bond. However, Tables 1 and 2 show that this is not strictly correct. They show that
the CDSBond basis can be either positive or negative for par bond depending on the
interest rate curve shape.
3. Numerical Examples
We now show some pricing examples. The interest rate curve is the swap zero curve for
July 16, 2008 taken from Bloomberg. The payment frequency is semiannual. We linearly
7
interpolate the swap zero curve to obtain the zero rates for all payment dates. For our
purpose, linear interpolation is deemed adequate because we only need a forward Libor
curve. However, different interpolation scheme may and will result in slightly different
bond implied CDS spread. The impact of interpolation scheme on ASW spread seems to
be smaller than on CDS spread.
Let
k
Z be the zero rate for maturity
k
T , the forward Libor rate for period ( )
k k
T T ,
1 −
is
( )
1
1 1
1 1
, 0
−
− −
− −
−
−
= =
k k
k k k k
k k
T T
Z T Z T
T L L (9)
Table 1: Bond implied CDS spread and ASW spread as function of bond price 1D using
method I. The bond has 10 year to maturity with 7% semiannual coupon, 40% recovery.
The day count convention is 30/360. Note that all values are in percentage except for W.
D 10 5 0 5 10 15 20
W 1.06 1.03 1.0 0.96 0.93 0.88 0.83
CDS 0.96 1.60 2.31 3.09 3.97 4.97 6.13
ASW 1.05 1.67 2.29 2.92 3.54 4.16 4.79
Basis 0.09 0.07 0.01 0.17 0.43 0.81 1.34
λ 1.58 2.64 3.80 5.09 6.65 8.19 10.11
Risky
L L −
0.03 0.05 0.08 0.11 0.14 0.17 0.21
( ) 01 / 01 1 PV PV C −
0.03 0.05 0.07 0.09 0.11 0.14 0.18
A D PV D / 01 / −
0.09 0.07 0 0.15 0.40 0.78 1.31
Table 1 shows the bond implied CDS spread (formula (7)), ASW spread (formula (1))
and CDS basis (formula (8) as a function of the bond price discount D. The bond pays
7% coupon semiannually and has 10 years to maturity. We assume 40% recovery rate
and 30/360 day count convention. The CDS spread, ASW spread and bond implied
hazard rate all increase with decreasing bond price (increasing D). As expected, the CDS
Bond basis decreases with increasing bond price.
Table 1 also shows the three terms in the CDSBond basis formula (7). The individual
contribution of the three terms in formula (7) can be easily inferred from Table 1. In this
case, the first term slightly dominates the second term, and the net effect is small due to
offsetting. But this is not always the case. For example, the first term in (7) is zero if the
Libor curve is flat. For bond trading substantially away from par, the 3
rd
term in formula
(7) dominates the basis.
To demonstrate the curve effect, Table 2 shows the results for the same bond as in Table
1 but with a flat Libor curve of 4.7% which is the average Libor in Table 1. We can see
that the shape of interest rate curve has a small effect on the spreads and basis.
8
Table 2: Flat Libor curve. The same bond as in Table 1. All values are in percentage.
D 10 5 0 5 10 15 20
CDS 0.92 1.55 2.24 3.00 3.86 4.83 5.95
ASW 1.04 1.67 2.30 2.93 3.56 4.20 4.83
Basis 0.12 0.12 0.06 0.07 0.29 0.63 1.12
λ 1.51 2.55 3.68 4.94 6.35 7.96 9.81
Risky
L L −
0.00 0.00 0.00 0.00 0.00 0.00 0.00
( ) 01 / 01 1 PV PV C −
0.03 0.04 0.06 0.09 0.11 0.14 0.17
A D PV D / 01 / −
0.09 0.08 0.00 0.15 0.4 0.77 1.30
4. Conclusions
We have presented a simple explicit formula to calculate the CDS spread implied by the
bond market price. The value of the model is that it can be used either for issuers having
a single bond outstanding or issuers having multiple bond issues. The formula explicitly
expresses the bond implied CDS spread as the weighted sum of three factors: bond
coupon, bond discount percentage and the Libor curve.
A potential use of the spread formula (7) is to explore the difference between the market
CDS spread quote and the fair bond implied CDS spread.
5. References
[1] R. Pullirsch, R. Jankowitsch, T. Veza, The Delivery Option in Credit Default Swaps,
Working paper, October 25, 2007.
[2] J. Hull, M. Predescu, A.White, The Relationship Between Credit Default Swap
Spreads, Bond Yields, and Credit Rating Announcements, Journal Banking and
Finance, V28, pp 27892811, 2004.
[3] M. Davies, D. Pugachevsky, Bond spreads as a proxy for credit default swap spreads,
Risk magazine, 2005.
[4] A. Berd, R. Mashal, P. Wang, Defining, Estimating and Using Credit Term Structure,
Part 1, 2, 3, November 2004.
[5] D. Lando, On Cox processes and credit risky securities, working paper, 1998.
[6] X. Guo, R. Jarrow, C. Menn, A Note on Lando’s Formula and Conditional
Independence, working paper May, 2007.
[7] D. O’kane, R. McAdie, Explaining the Basis: Cash versus Default Swaps, Lehman
Brothers Report, May 2001.
9
Appendix A
We derive the pricing formula (4) for the bond implied CDS spread. As stated previously,
given the bond discount D, the negative trade investor buys the bond and hedge with
buying 1D/(1R) notional CDS protection. This CDS notional amount makes the
combined CDSbond position default neutral.
Suppose the investor funds the purchase at Libor flat which is a reasonable assumption
because CDS spread is based on Libor [2]. We assume that the cash flow terminates upon
default. Furthermore, we assume the investor will pay the accrued CDS premium and
loan interest but not receive bond accrued coupon when the bond defaults. Based on these
assumptions, the cash flow to the investor is described in the following table.
CDS Loan Bond Total
Initial 0 1D (1D) 0
Payment Date (1D/(1R))S (1D)L C C(1D)L 
(1D/(1R))S
Default (1DR)Accrued
premium
(1D)accrued
loan interest
R Accrued CDS and
loan interest
Maturity 0 (1D) 1 D
The noarbitrage condition means that the expected present value of all cash flow, initial
and future, to the investor must be zero. We arrive at
( ) ( ) ( )
( )
( ) ( )
( )
( )
( )
( )
0
1
1
1
1 , 1
1
1
1 , 1
1
1
1
1 1
1
1 1 1
0
=
)
`
¹ >
+
≤ < ×
−
×
(
¸
(
¸

¹

\

−
− + − −
¦
¦
¹
¦
¦
´
¦
(
(
(
(
¸
(
¸
> ∆


¹

\


¹

\

−
− − − −
−
−
=
− −
=
− − −
∑
∑
N
N
k k
k
N
k
CDS k k
N
k k
k k CDS k k
T B
T
D
T T
B
T
S
R
D
T T L D
T B
T T S
R
D
T T L D C
E
τ
τ
τ
τ
τ
(A.1)
In the above equation, ( ) T t L , is the Tmaturity forward Libor seen at time t, R is the expected
recovery rate,τ is the default time, B(t) is the money market account and 1(A) is the indicator
function. The first term in equation (A.1) is the net coupon payment to the investor on
scheduled payment dates. The second term is the payment upon default of accrued loan
interest and CDS spread since the last scheduled coupon payment date. We assume the
accrued bond coupon is not paid upon default. The third term is the payoff to the investor
at maturity if the bond has not defaulted.
Adopting the usual assumption of independence between credit spread and interest rate,
using the Lando formula (see [5]), and approximating the integral using trapezoidal rule,
we get
10
( )
( ) ( )
( ) ( ) ( )
( ) ( )
( ) ( )  
k k
k k
T
T
T
T
I k k
T P T P
T DF T DF
t dP t DF t dP
t B
E
B
T T
E
k
k
k
k
> − >
+
=
> − = >


¹

\

− =
)
`
¹
¹
´
¦ ≤ <
−
−
−
∫ ∫
− −
τ τ
τ τ
τ
τ
1
1
1
0
2
1 1
1 1
(A.2)
( ) ( )
( ) ( )
( )
( ) ( ) ( ) ( ) ( ) ( )  
k k k k
T
T
k
T
T
k I k k k
T P T P T DF T t dP t DF T t
t dP
t B
T t
E
B
T T T
E
k
k
k
k
> − > ∆ = > − − =
>


¹

\
 −
− =
)
`
¹
¹
´
¦ ≤ < −
− − −
− − −
∫
∫
−
−
τ τ τ
τ
τ
τ τ
1 1 1
1 1 1
0
2
1
1
1
1
(A.3)
Using the fact that the ( )
1
,
− k
T t L is a martingale under the −
k
T forward measure, we have
( )
( ) ( )
( )
( )
( )
( )
( ) ( ) ( ) ( ) ( )  
( ) ( ) ( )  
k k k k k
k k k k
T
k k
T
T k
k
k k
I k k k
k k
T P T P T DF L T
T P T P T T L E T DF T
t dP
T B
T
T T L E
B
T T T
T T L E
k
k
k
> − > ∆ =
> − > ∆ =
>


¹

\
 ∆
− =
)
`
¹
¹
´
¦ ≤ < −
− − −
− − − −
−
− −
− −
− −
∫
−
τ τ
τ τ
τ
τ
τ τ
1 1 1
1 1 1 1
1
1 1
1 1
1 1 0
2
1
,
2
1
,
2
1 1
,
1
(A.4)
where ( )
1 1
, 0
− −
=
k k
T L L is the −
−1 k
T forward rate.
Substituting equations (A.2), (A.3) and (A.4) into (A.1), and rearranging terms results in
( ) ( ) ( ) 0 01
1
1 1 01 = > × +
(
¸
(
¸

¹

\

−
− + − − ×
N N CDS
Risky
T P T DF D PV S
R
D
L D PV C τ . (A.5)
where
( ) ( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( ) ( )
( ) ( )
∏
∫
∑
∑
∑
∑
= − −
=
− −
=
− − −
=
− −
=
−
∆ +
=


¹

\

− = >
> − > + =
> + > ∆ =
> + > ∆ =
> ∆ =
k
j j j
k
T
k
N
k
k k k k
N
k
k k k k k
Risky
N
k
k k k k
N
k
k k k
T L
T DF dt Exp T P
T P T P T DF T DF Loss
PV T P T P T DF T L L
T P T P T DF T PV
T P T T DF PV
k
1 1 1 0
1
1 1
1
1 1 1
1
1 1
1
1
1
1
,
2
1
01 /
2
1
2
1
01
, 01
λ τ
τ τ
τ τ
τ τ
τ
(A.5)
11
Appendix B
Now we prove the equivalency between equations (4) and (6) under the constraint of
equation (5).
Proposition: If equation (5) is satisfied, the bond implied CDS spread calculated using
equation (4) satisfies equation (6).
Proof: By virtue of equation (5), we can rewrite equation (4) as
( ) 0 01
1
01 1 = × −
−
×
− − − PV S W
D
Loss R
PV L P DF
CDS
Risky
N N
(B.1)
where ( )
N N
T DF DF = , ( )
N N
T P P > = τ .
Furthermore, we rewrite the term Loss in equation (A.1) into
( )( )    
∑ ∑
=
− −
=
− −
− + − = − + =
N
k
k k k k N N
N
k
k k k k
P DF P DF P DF P P DF DF Loss
1
1 1
1
1 1
2
1
1
2
1
2
1
(B.2)
Notice that
k k k k k
DF DF T L DF − = ∆
− − − 1 1 1
and using (B.2), we have
( ) ( )
( )( ) ( )  
( )( )
∑
∑
∑
=
− −
=
− − − −
=
− −
= − = − + − =
− + + − − =
− − − = − −
N
k
k k k k
N
k
k k k k k k k k
N
k
k k k k
Risky Risky
N N
Loss Loss Loss P P DF DF Loss
P DF P DF P P DF DF Loss
P DF P DF PV L Loss PV L P DF
1
1 1
1
1 1 1 1
1
1 1
2
2
1
2
2
2
1
2
01 2 01 1
(B.3)
Substitute (B.3) into (B.1) yields
0 01
1
1 = − × 
¹

\

−
− PV S
W
Loss
D
R
CDS
(B.4)
Since ( ) D
R
D
W − 
¹

\

−
− = 1 /
1
1 , we have . 1 /
1
1 R W
D
R
− = 
¹

\

−
− This completes the proof.
we describe a simple model for estimating the bond implied CDS spread. recovery rate. the basis is positive. Section 3 gives some numerical examples. Berd et al proposed to use the survivalbased modeling as a consistent measure for creditrisky bond pricing and risk management [4].e. An asset swap (ASW) is a package transaction between two parties in which the ASW buyer purchases a bond from the other party and simultaneously enters into an interest rate swap transaction. In this paper. but an explicit formula that combines all these factors seems needed and this paper provide a such formula. bond coupon and the price deviation from par on the credit spread. In a series of paper. We define the bond implied CDS spread as the spread that equates the bond market price with the bond fair theoretical price. the CDS is generally considered to be more attractive than the bond. The detailed model derivation is given in Appendices A and B. CDS and Asset Swap Spreads In this section. Section 2 describes the underlying theory and main formulas. to exchange the coupon on the bond for Libor plus a spread. the Libor curve is implicitly an AA rated yield curve. The resulting survival probability term structure is not necessarily the same as that implied by the CDS market. i. bond price. The reverse is true if the basis is negative. A common asset swap is the par asset swap where the buyer pays par at the inception of the deal. The model concept is not new as it is based on survival probability.are the borrowing rates between banks of AA rating. In their model. we use Libor as the risk free interest rate. Bond implied CDS spreads have been previously investigated. These effects have been discussed in literature. and interest rate and credit curves. usually with the same counterparty. CDSBond basis is the difference between the CDS spread and the ASW spread on the same bond. duration. As a result. For example. convexity and recovery rate [3]. Davies and Pugachevsky proposed an approximation method for calculating the bond implied CDS spread based on the Z spread adjusted by the bond’s market price. The paper is organized as follows. It is a general indicator of relative value of CDS versus the cash bond. The spread is called the asset swap spread. we describe a model for calculating the bond implied CDS spread and the CDSBond basis. The idiosyncratic fitting error is accounted for by the OAStofit which is essentially a measure in spread of aggregate effect of market factors on the bond. when the CDS spread is higher than the ASW spread. We demonstrate the effects of Libor curve. and section 4 concludes the paper. But it is cast in such an explicit way that it is easy to analyze the effects on the CDS spread of bond coupon. 2 . Unlike CDS. 2. ASW continues following bond default. the survival probability term structure of an issuer is estimated by regressing model prices against market prices across all bonds of that issuer under a constant recovery rate.
This discount amount is paid upfront by the investor and recouped over the life of the swap as part of the swap spread. 1D is the bond’s dirty price. 3 . the investor buys a package consisting of a cash bond and a payer interest swap where the investor swaps the bond coupon for Libor plus a (ASW) spread. The spread of a par asset swap is given by S ASW = M − (1 − D) D =C−L + A A (1) where M is the bond riskfree price. This means that the bond dirty price and the initial swap value must sum to par.2. Let T0 be today. For a discount bond. and the bond coupon payment dates be Tk . It states that the ASW spread is the difference between the riskless price and the market price amortized over the swap life. Furthermore. we assume that the ASW payment dates coincide with the bond coupon payment dates. The market practice is asset swap does not knock out when the underlying bond defaults. Since the bond riskfree price is always greater than the market price. ∆Tk −1 = Tk − Tk −1 . k = 1. As a result.. For a premium bond..1 Asset Swap Spread In a par asset swap transaction. ASW spread is always positive. but more intuitive. the initial swap value is positive to the investor.. and A is the riskfree annuity given by A = ∑ ∆Tk −1 DF (Tk ) k =1 N (2) where DF (Tk ) is the discount factor for maturity Tk . the average Libor over the swap life and the difference of bond dirty price and par scaled by the riskless annuity. and the riskfree average Libor weighted by the discount factors is L= 1 N ∑ Lk −1∆Tk −1 DF (Tk ) A k =1 (3) The first form of ASW spread in formula (1) is well known. the investor bears some interest rate risk in that he needs to seek new funding to pay the fixed rate should the bond default. The second form of formula (1) is less familiar. Lk −1 is the Tk −1 − maturity forward Libor rate paid at Tk . The last component can also be interpreted as the bond discount amount amortized over the swap life. it is negative. It shows that the ASW spread is composed of three components: the bond coupon. The price of the asset swap package is par (hence the term par asset swap) which the investor pays at the deal inception. C is the bond coupon. Note that the spread would have to be negative if the bond market price is greater than the bond riskfree price. N ..
Remarks: 1) If the bond is trading at par (D = 0). Let us assume that the investor borrows 1D to purchase a bond with fixed coupon C. the more deeply discounted is the bond. the investor needs to buy 1D/(1R) notional CDS protection in order to make the combined position default neutral. 2) The ASW spread increases as the bond price decreases (D increases). Given a nonzero recovery. Assuming a recovery rate R. S CDS = 1 W PV 01 D R × Loss − L Risky + C 1 − (1 − D ) PV 01 PV 01 (4) D R D under the constraint where W = 1 − / (1 − D ) = 1 − 1− R 1− D 1− R 1 − D = C × PV 01 + R × Loss + P(τ > TN ) × DF (TN ) (5) Equation (4) shows that the bond implied CDS spread is the sum of contributions from the bond coupon. resulting in an increasing basis. and the recovery rate augmented by the risky annuities and W which is the ratio of the CDS notional amount to the bond price. And the bond implied spread increases with decreasing bond price (increasing D) but at a faster rate than ASW spread does. 3) We will show in the next section that the bond implied spreads – given by formula (7) – has characteristics similar to the ASW spread. the higher is the ASW spread. Denoting the recovery rate by R. 2. The market price of the bond is 1D where D represents bond’s discount relative to par.2 Bond Implied CDS Spread Suppose an investor executes a socalled negative basis trade in which he buys the cash bond on the ASW basis and buys CDS protection. He then buys 1D/(1R) notional CDS protection on the bond. (5) and (6)). the bond implied CDS spread is given by (See Appendix A for derivation and definitions of the terms in equations (4). the Libor curve. the ASW spread is the difference between the bond coupon and the average Libor rate over the ASW term. The bond implied spread contains three terms with interpretations similar to those of terms in formula (1). this amount is less (more) than 1D for a discount (premium) bond. We will show in Appendix B that the bond implied CDS spread calculated using equation (4) automatically satisfies the standard CDS pricing equation 4 . Hence. the difference between the bond (dirty) price and par.
the default probability implied by equation (6) is generally inconsistent with equation (5). Substituting equation (6) into equation (4) and solve for S CDS . 5 . 3) Term structures of market CDS spread and bond price: We bootstrap to calculate the hazard rate term structure and the bond implied recovery rate by simultaneously solving equations (5) and (6). we find a simplified form S CDS = C × PV 01 D − L Risky + PV 01 PV 01 (7) It is clear from formula (7) that for a par bond (D = 0) and a flat Libor curve. Given a CDS spread and a recovery rate R. the recovery rates are influenced by the cash market factors and are not necessarily the expected percentage recovery amount of par. the model framework of (57) can be used in several ways: 1) Single bond: In this situation. we bootstrap to obtain a term structure of the bond implied hazard rate consistent with the given bond prices. However. if we are concerned only with the CDS spread. The bond implied hazard rate is not necessarily consistent with the hazard rate implied by the market CDS quote of the same issuer. and calculate the constant bond implied hazard rate using equation (5). the bond accrued coupon is not paid but the accrued CDS premium and Libor interest are paid upon default. The difference is due to that in our model. is the model consistent with the standard CDS pricing model (6). equation (7) can be used to compare the bond implied CDS spread to the market CDS quote. The reverse does not hold. Therefore. while there is no such default payment discrepancy in the par floater replication. The bond implied CDS spread is then calculated using equation (7). The bond implied hazard rates are determined by the fundamental and technical factors in the cash market while the CDS hazard rates are determined by the CDS market fundamental and technical factors. if and only if the recovery and default probability are consistent with the bond price. However.S CDS = 1− R × Loss PV 01 (6) This implies that. The bond implied CDS spread for a maturity can then be calculated from (7). The resulting hazard rate and recovery rate term structures are consistent with the bond market prices and the market CDS spreads of the issuer. equation (7) is all that matters. we assume a recovery rate R. 2) Multiple bonds of differing maturities: Assuming a constant recovery. Given the bond’s market price 1D. we have PV 01 − L which is slightly less than the CL implied by the risky par floater S CDS = C PV 01 replication model.
or increasing default risk. the CDS spread formula (7) can also be directly obtained if the CDS notional is 1D. 3) It is important to note that in equations (47) we have not imposed any restriction on the shape of credit and interest rate curves. While the investor does not receive the bond’s accrued interest in the event of default. Tables 1 and 2 show that this is not strictly correct. 1D/(1R) CDS notional is default neutral. 2) The second term is due to the payment mismatch between the bond coupon and CDS premium and borrowing cost upon default. This term increases with increasing bond discount D. CDSBond basis is positive for discount bond and negative for premium bond. Therefore. 2008 taken from Bloomberg.Remarks: 1) Interestingly. allowing easy analysis of individual factors. and bond price augmented by the risky annuities. the first term can be interpreted as the impact of the yield curve slope to the basis. They show that the CDSBond basis can be either positive or negative for par bond depending on the interest rate curve shape. It is zero when the forward curve is flat. It is positive for discount bond and negative for premium bond. rough speaking. the interest rate curve effect on the basis is generally positive (see Table 1). It is positive for an upward sloping interest rate curve. This term always contributes negatively to the CDSBond basis. and negative for a downward sloping curve. However. Numerical Examples We now show some pricing examples. he still needs to pay the accrued CDS premium and interest on the loan. Since the normal yield curve shape is upward sloping. We linearly 6 . 1D notional would result in a small default payoff of DR which is offset by the larger carry (1D)*S.3 CDS–Bond Basis The CDS–Bond basis is defined as the difference between the bond implied CDS spread (7) and the par asset swap spread (1) D PV 01 PV 01 1 − Basis = L − L Risky − C 1 − + A PV 01 PV 01 ( ) (8) The CDSBond basis consists of three terms: 1) The first term is the difference between the average riskfree Libor rate and the average risky Libor rate. 2. 2) The form of equation (7) is useful as it explicitly expresses the CDS spread in terms of bond coupon. The payment frequency is semiannual. 3) The third term represents the effect of bond’s market price on the CDSBond basis. It explains why. 3. The interest rate curve is the swap zero curve for July 16. interest rate curve.
31 Table 1 shows the bond implied CDS spread (formula (7)).93 0.58 2. We assume 40% recovery rate and 30/360 day count convention. different interpolation scheme may and will result in slightly different bond implied CDS spread.7% which is the average Libor in Table 1.17 0. ASW spread (formula (1)) and CDS basis (formula (8) as a function of the bond price discount D.17 0.interpolate the swap zero curve to obtain the zero rates for all payment dates. The bond pays 7% coupon semiannually and has 10 years to maturity.60 2.21 L −L C 1 − PV 01 / PV 01 ( ) 0.07 0. For bond trading substantially away from par.03 0. For example.78 1. the forward Libor rate for period (Tk −1 .97 4.88 0. D 10 5 0 5 10 15 20 W 1.43 0.29 2. But this is not always the case. We can see that the shape of interest rate curve has a small effect on the spreads and basis. In this case.07 0 0.79 Basis 0.80 5.05 1. the first term slightly dominates the second term.65 8.07 0.31 3.54 4.11 0.34 λ 1.97 6. ASW spread and bond implied hazard rate all increase with decreasing bond price (increasing D).15 0. The bond has 10 year to maturity with 7% semiannual coupon.81 1.01 0. and the net effect is small due to offsetting.64 3. The impact of interpolation scheme on ASW spread seems to be smaller than on CDS spread.67 2.16 4.09 0. Let Z k be the zero rate for maturity Tk .09 0. the 3rd term in formula (7) dominates the basis. the first term in (7) is zero if the Libor curve is flat.83 CDS 0. linear interpolation is deemed adequate because we only need a forward Libor curve. Tk −1 ) = Tk Z k − Tk −1 Z k −1 Tk − Tk −1 (9) Table 1: Bond implied CDS spread and ASW spread as function of bond price 1D using method I.09 6. As expected. The CDS spread. The individual contribution of the three terms in formula (7) can be easily inferred from Table 1. For our purpose.11 Risky 0.92 3.40 0. Table 2 shows the results for the same bond as in Table 1 but with a flat Libor curve of 4.08 0. To demonstrate the curve effect.14 0.06 1. The day count convention is 30/360. 40% recovery.09 0.96 0.13 ASW 1.96 1.0 0.05 0. the CDSBond basis decreases with increasing bond price. Tk ) is Lk −1 = L(0.09 3.14 0.05 0.03 0.03 1.19 10.18 D / PV 01 − D / A 0. However.11 0. Note that all values are in percentage except for W. Table 1 also shows the three terms in the CDSBond basis formula (7). 7 .
51 2. References [1] R. [4] A.12 0. A potential use of the spread formula (7) is to explore the difference between the market CDS spread quote and the fair bond implied CDS spread. October 25. C. Veza. Berd. Conclusions We have presented a simple explicit formula to calculate the CDS spread implied by the bond market price. M.86 4. 2004. pp 27892811.55 3.77 1.09 0.94 6.00 L −L C 1 − PV 01 / PV 01 ( ) 0. On Cox processes and credit risky securities.06 0.06 0.00 0.56 4. Working paper.95 ASW 1.35 7. A. Part 1. P. Jarrow. The formula explicitly expresses the bond implied CDS spread as the weighted sum of three factors: bond coupon. All values are in percentage.00 0.30 2. Journal Banking and Finance. Bond Yields. Davies.68 4. [2] J. November 2004. Menn. T.15 0.17 D / PV 01 − D / A 0. Estimating and Using Credit Term Structure.67 2.04 0. R.24 3.30 4.00 0. R.04 1.83 5. Guo. [7] D. The Delivery Option in Credit Default Swaps.00 0. O’kane. R. Predescu.03 0.08 0.12 λ 1. Defining. working paper May.07 0. A Note on Lando’s Formula and Conditional Independence. 2.81 Risky 0.29 0. May 2001.00 3.96 9. Hull.00 0. 1998.63 1.20 4. Risk magazine.92 1. 3. The Relationship Between Credit Default Swap Spreads. Pullirsch. Lehman Brothers Report. 2005. [3] M. McAdie. D. R.00 0.14 0.00 0. The value of the model is that it can be used either for issuers having a single bond outstanding or issuers having multiple bond issues.09 0. Explaining the Basis: Cash versus Default Swaps. The same bond as in Table 1.11 0. and Credit Rating Announcements. Lando.White. 5. Wang.83 Basis 0. Mashal.93 3. bond discount percentage and the Libor curve. working paper. 8 . 2007. D 10 5 0 5 10 15 20 CDS 0. V28. [6] X.55 2. 2007.Table 2: Flat Libor curve. Bond spreads as a proxy for credit default swap spreads. Pugachevsky.4 0. [5] D. Jankowitsch.12 0.
τ is the default time. We arrive at D S CDS ∆Tk −11(τ > Tk ) N C − (1 − D )L(Tk −1 .1) In the above equation. Adopting the usual assumption of independence between credit spread and interest rate. Tk −1 ) − 1 − 1− R E 0 ∑ B(Tk ) k =1 N τ − Tk −1 D − ∑ (1 − D )L(Tk −1 . we assume the investor will pay the accrued CDS premium and loan interest but not receive bond accrued coupon when the bond defaults. B(t) is the money market account and 1(A) is the indicator function. The second term is the payment upon default of accrued loan interest and CDS spread since the last scheduled coupon payment date. given the bond discount D. This CDS notional amount makes the combined CDSbond position default neutral. to the investor must be zero. Suppose the investor funds the purchase at Libor flat which is a reasonable assumption because CDS spread is based on Libor [2]. we get 9 . L(t . Based on these assumptions. initial and future. CDS 0 (1D/(1R))S (1DR)Accrued premium 0 Loan 1D (1D)L (1D)accrued loan interest (1D) Bond (1D) C R 1 Total 0 C(1D)L (1D/(1R))S Accrued CDS and loan interest D Initial Payment Date Default Maturity The noarbitrage condition means that the expected present value of all cash flow. the negative trade investor buys the bond and hedge with buying 1D/(1R) notional CDS protection. Furthermore. T ) is the Tmaturity forward Libor seen at time t. the cash flow to the investor is described in the following table. As stated previously. The first term in equation (A. We assume the accrued bond coupon is not paid upon default.1) is the net coupon payment to the investor on scheduled payment dates. R is the expected recovery rate. The third term is the payoff to the investor at maturity if the bond has not defaulted.Appendix A We derive the pricing formula (4) for the bond implied CDS spread. using the Lando formula (see [5]). and approximating the integral using trapezoidal rule. Tk −1 ) + 1 − × 1(Tk −1 < τ ≤ Tk ) S CDS × B(τ ) 1− R k =1 +D 1(τ > TN ) =0 B(TN ) (A. We assume that the cash flow terminates upon default.
we have (τ − Tk −1 )1(Tk −1 < τ ≤ Tk ) 1 I ∆Tk −1 k dP(τ > t ) E 0 L(Tk −1 .5) PV 01 = L Risky = Loss = 1 N ∑ ∆Tk −1 DF (Tk )(P(τ > Tk −1 ) + P(τ > Tk )) 2 k =1 1 N ∑ Lk −1∆Tk −1 DF (Tk )(P(τ > Tk −1 ) + P(τ > Tk )) / PV 01 2 k =1 (A.2).4) where Lk −1 = L(0. (A.DF (Tk −1 ) + DF (Tk ) [P(τ > Tk −1 ) − P(τ > Tk )] = 2 Tk t − Tk −1 (τ − Tk −1 )1(Tk −1 < τ ≤ Tk ) E0 = − ∫ EI B(t ) dP(τ > t ) B(τ ) Tk −1 Tk k k 1(Tk −1 < τ ≤ Tk ) I 1 dP(τ > t ) = − ∫ DF (t )dP(τ > t ) E0 =− ∫ E B(τ ) B(t ) Tk −1 Tk −1 T T (A.1).3) and (A. Substituting equations (A. Tk −1 ) = − E L(Tk −1 . k =1 N (A. Tk −1 ))[P(τ > Tk −1 ) − P(τ > Tk )] 2 1 = ∆Tk −1 Lk −1 DF (Tk )[P(τ > Tk −1 ) − P(τ > Tk )] 2 (A.3) 1 = − ∫ (t − Tk −1 )DF (t )dP(τ > t ) = ∆Tk −1 DF (Tk )[P(τ > Tk −1 ) − P(τ > Tk )] 2 Tk −1 Using the fact that the L(t . 1− R where PV 01 = ∑ DF (Tk )∆Tk −1 P(τ > Tk ).5) 1 N ∑ (DF (Tk −1 ) + DF (Tk ))(P(τ > Tk −1 ) − P(τ > Tk )) 2 k =1 DF (Tk ) = ∏ j =1 k Tk P(τ > Tk ) = Exp − ∫ λdt . and rearranging terms results in D C × PV 01 − (1 − D )L Risky + 1 − S CDS PV 01 + D × DF (TN )P(τ > TN ) = 0 .2) (A.4) into (A. Tk −1 ) is a martingale under the Tk − forward measure. Tk −1 ) is the Tk −1 − forward rate. Tk −1 ) 2 B(Tk ) Tk∫−1 B(τ ) T = 1 ∆Tk −1 DF (Tk )E Tk (L(Tk −1 . 0 1 1 + L j −1 ∆T j −1 10 .
we have (1 − DFN PN ) − L Risky PV 01 = 2 Loss − L Risky PV 01 − ∑ (DFk Pk −1 − DFk −1 Pk ) k =1 N = 2 Loss − = 2 Loss − 1 ∑ [(DFk −1 − DFk )(Pk −1 + Pk ) + 2(DFk Pk −1 − DFk −1 Pk )] 2 k =1 1 N ∑ (DFk −1 + DFk )(Pk −1 − Pk ) = 2Loss − Loss = Loss 2 k =1 N (B. Proposition: If equation (5) is satisfied. we can rewrite equation (4) as (1 − DFN PN ) − L Risky PV 01 − R × Loss − W × S CDS PV 01 = 0 1− D (B.3) into (B. Proof: By virtue of equation (5). we rewrite the term Loss in equation (A. Furthermore.1) yields R Loss (B.1) where DFN = DF (TN ) . we have 1 − / W = 1 − R. the bond implied CDS spread calculated using equation (4) satisfies equation (6).3) Substitute (B.2). PN = P(τ > TN ) .4) − S CDS PV 01 = 0 1 − × 1− D W D R Since W = 1 − / (1 − D ) .2) ∑ 2 k =1 2 2 k =1 Notice that DFk Lk −1 ∆Tk −1 = DFk −1 − DFk and using (B.Appendix B Now we prove the equivalency between equations (4) and (6) under the constraint of equation (5). This completes the proof.1) into Loss = N 1 N (DFk −1 + DFk )(Pk −1 − Pk ) = 1 [1 − DFN PN ] + 1 ∑ [DFk Pk −1 − DFk −1 Pk ] (B. 1− R 1− D 11 .