Valuation of Credit Default Swaptions and Credit Default Index Swaptions Nice - Rutkowski

Credit Default Swaptions
Credit Default Index Swaptions

Market Models for CDS Spreads
Valuation of Credit Default Swaptions
and Credit Default Index Swaptions
Marek Rutkowski
School of Mathematics and Statistics
University of New South Wales
Sydney, Australia
Recent Advances in the Theory and Practice of Credit Derivatives
CNRS and University of Nice Sophia Antipolis
September 28-30, 2009
M. Rutkowski Credit Default Swaps and Swaptions
Outline
1
Credit Default Swaps and Swaptions
2
Hazard Process Approach
3
Market Pricing Formulae
4
CIR Default Intensity Model
5
6
References on Valuation of Credit Default Swaptions
D. Brigo and M. Morini: CDS market formulas and models. Working
paper, Banca IMI, 2005.
D. Brigo and A. Alfonsi: Credit default swaps calibration and option
pricing with the SSRD stochastic intensity and interest-rate model.
Finance and Stochastics 9 (2005), 29-42.
F. Jamshidian: Valuation of credit default swaps and swaptions.
Finance and Stochastics 8 (2004), 343371.
M. Morini and D. Brigo: No-armageddon arbitrage-free equivalent
measure for index options in a credit crisis. Working paper, Banca IMI
and Fitch Solutions, 2007.
M. Rutkowski and A. Armstrong: Valuation of credit default swaptions
and credit default index swaptions. Working paper, UNSW, 2007.
References on Modelling of CDS Spreads
N. Bennani and D. Dahan: An extended market model for credit
derivatives. Presented at the international conference Stochastic
Finance, Lisbon, 2004.
D. Brigo: Candidate market models and the calibrated CIR++ stochastic
intensity model for credit default swap options and callable oaters.
In: Proceedings of the 4th ICS Conference, Tokyo, March 18-19, 2004.
D. Brigo: Constant maturity credit default swap pricing with market
models. Working paper, Banca IMI, 2004.
L. Li and M. Rutkowski: Market models for forward swap rates and
forward CDS spreads. Working paper, UNSW, 2009.
L. Schlgl: Note on CDS market models. Working paper, Lehman
Brothers, 2007.
References on Hedging of Credit Default Swaptions
T. Bielecki, M. Jeanblanc and M. Rutkowski: Hedging of basket credit
derivatives in credit default swap market. Journal of Credit Risk 3 (2007),
91-132.
T. Bielecki, M. Jeanblanc and M. Rutkowski: Pricing and trading credit
default swaps in a hazard process model. Annals of Applied Probability
18 (2008), 2495-2529.
T. Bielecki, M. Jeanblanc and M. Rutkowski: Valuation and hedging of
credit default swaptions in the CIR default intensity model. Working
paper, 2008.
Valuation of Forward Credit Default Swaps
Hedging of Credit Default Swaptions
Hazard Process Set-up
Terminology and notation:
1
The default time is a strictly positive random variable dened on the
underlying probability space (, G, P).
2
We dene the default indicator process H
t
= 1
{t }
and we denote by H
its natural ltration.
3
We assume that we are given, in addition, some auxiliary ltration F and
we write G = H F, meaning that G
t
= (H
t
, F
t
) for every t R
+
.
4
The ltration F is termed the reference ltration.
5
The ltration G is called the full ltration.
Martingale Measure
The underlying market model is arbitrage-free, in the following sense:
1
Let the savings account B be given by
B
t
= exp
_
_
t
0
r
u
du
_
, t R
+
,
where the short-term rate r follows an F-adapted process.
2
A spot martingale measure Q is associated with the choice of the
savings account B as a numraire.
3
The underlying market model is arbitrage-free, meaning that it admits a
spot martingale measure Q equivalent to P. Uniqueness of a martingale
measure is not postulated.
Hazard Process
Let us summarize the main features of the hazard process approach:
1
Let us denote by
G
t
= Q( > t | F
t
)
the survival process of with respect to the reference ltration F.
We postulate that G
0
= 1 and G
t
> 0 for every t [0, T].
2
We dene the hazard process = ln G of with respect to the
ltration F.
3
For any Q-integrable and F
T
-measurable random variable Y, the
following classic formula is valid
E
Q
(1
{T<}
Y | G
t
) = 1
{t <}
G
1
t
E
Q
(G
T
Y | F
t
).
Default Intensity
1
Assume that the supermartingale G is continuous.
2
We denote by G = its Doob-Meyer decomposition.
3
Let the increasing process be absolutely continuous, that is,
d
t
=
t
dt for some F-adapted and non-negative process .
4
Then the process
t
= G
1
t

t
is called the F-intensity of default time.
Lemma
The process M, given by the formula
M
t
= H
t

_
t
0
u
du = H
t

_
t
0
(1 H
u
)
u
du,
is a (Q, G)-martingale.
Defaultable Claim
A generic defaultable claim (X, A, Z, ) consists of:
1
A promised contingent claim X representing the payoff received by the
holder of the claim at time T, if no default has occurred prior to or at
maturity date T.
2
A process A representing the dividends stream prior to default.
3
A recovery process Z representing the recovery payoff at time of default,
if default occurs prior to or at maturity date T.
4
A random time representing the default time.
Denition
The dividend process D of a defaultable claim (X, A, Z, ) maturing at T
equals, for every t [0, T],
D
t
= X1
{>T}
1
[T,[
(t ) +
_
]0,t ]
(1 H
u
) dA
u
+
_
]0,t ]
Z
u
dH
u
.
Ex-dividend Price
Recall that:
The process B represents the savings account.
A probability measure Q is a spot martingale measure.
Denition
The ex-dividend price S associated with the dividend process D equals,
for every t [0, T],
S
t
= B
t
E
Q
_
_
]t ,T]
B
1
u
dD
u
G
t
_
= 1
{t <}

S
t
where Q is a spot martingale measure.
The ex-dividend price represents the (market) value of a defaultable
claim.
The F-adapted process

S is termed the pre-default value.
Valuation Formula
Lemma
The value of a defaultable claim (X, A, Z, ) maturing at T equals
S
t
= 1
{t <}
B
t
G
t
E
Q
_
B
1
T
G
T
X1
{t <T}
+
_
T
t
B
1
u
G
u
Z
u
u
du+
_
T
t
B
1
u
G
u
dA
u
F
t
_
where Q is a martingale measure.
Recall that is the martingale part in the Doob-Meyer decomposition
of G.
Let m be the (Q, F)-martingale given by the formula
m
t
= E
Q
_
B
1
T
G
T
X +
_
T
0
B
1
u
G
u
Z
u
u
du +
_
T
0
B
1
u
G
u
dA
u
F
t
_
.
Price Dynamics
Proposition
The dynamics of the value process S on [0, T] are
dS
t
= S
t
dM
t
+ (1 H
t
)
_
(r
t
S
t

t
Z
t
) dt + dA
t
_
+ (1 H
t
)G
1
t
_
B
t
dm
t
S
t
d
t
_
+ (1 H
t
)G
2
t
_
S
t
d
t
B
t
d, m
t
_
.
The dynamics of the pre-default value

S on [0, T] are
d
S
t
=
_
(
t
+ r
t
)
S
t

t
Z
t
_
dt + dA
t
+ G
1
t
_
B
t
dm
t

S
t
d
t
_
+ G
2
t
_
S
t
d
t
B
t
d, m
t
_
.
Forward Credit Default Swap
Denition
A forward CDS issued at time s, with start date U, maturity T, and recovery
at default is a defaultable claim (0, A, Z, ) where
dA
t
= 1
]U,T]
(t ) dL
t
, Z
t
=
t
1
[U,T]
(t ).
An F
s
-measurable rate is the CDS rate.
An F-adapted process L species the tenor structure of fee payments.
An F-adapted process : [U, T] R represents the default protection.
Lemma
The value of the forward CDS equals, for every t [s, U],
S
t
() = B
t
E
Q
_
1
{U<T}
B
1
G
t
_
B
t
E
Q
_
_
]t U,T]
B
1
u
dL
u
G
t
_
.
Valuation of a Forward CDS
Lemma
The value of a credit default swap started at s, equals, for every t [s, U],
S
t
() = 1
{t <}
B
t
G
t
E
Q
_
_
T
U
B
1
u

u
dG
u

_
]U,T]
B
1
u
G
u
dL
u
F
t
_
.
Note that S
t
() = 1
{t <}
S
t
() where the F-adapted process

S() is the
pre-default value. Moreover
S
t
() =

P(t , U, T)
A(t , U, T)
where
P(t , U, T) is the pre-default value of the protection leg,
A(t , U, T) is the pre-default value of the fee leg per one unit of .
Forward CDS Rate
The forward CDS rate is dened similarly as the forward swap rate for
a default-free interest rate swap.
Denition
The forward market CDS at time t [0, U] is the forward CDS in which the
F
t
-measurable rate is such that the contract is valueless at time t .
The corresponding pre-default forward CDS rate at time t is the unique
F
t
-measurable random variable (t , U, T), which solves the equation
S
t
((t , U, T)) = 0.
Recall that for any F
t
-measurable rate we have that
S
t
() =

P(t , U, T)
A(t , U, T).
Forward CDS Rate
Lemma
For every t [0, U],
(t , U, T) =
P(t , U, T)
A(t , U, T)
=
E
Q
_
_
T
U
B
1
u

u
dG
u
F
t
_
E
Q
_
_
]U,T]
B
1
u
G
u
dL
u
F
t
_ =
M
P
t
M
A
t
where the (Q, F)-martingales M
P
and M
A
are given by
M
P
t
= E
Q
_
_
T
U
B
1
u

u
dG
u
F
t
_
and
M
A
t
= E
Q
_
_
]U,T]
B
1
u
G
u
dL
u
F
t
_
.
Credit Default Swaption
Denition
A credit default swaption is a call option with expiry date R U and zero
strike written on the value of the forward CDS issued at time 0 s < R,
with start date U, maturity T, and an F
s
-measurable rate .
The swaptions payoff C
R
at expiry equals C
R
= (S
R
())
+
.
Lemma
For a forward CDS with an F
s
-measurable rate we have, for every t [s, U],
S
t
() = 1
{t <}
A(t , U, T)((t , U, T) ).
It is clear that
C
R
= 1
{R<}
A(R, U, T)((R, U, T) )
+
.
A credit default swaption is formally equivalent to a call option on the forward
CDS rate with strike . This option is knocked out if default occurs prior to R.
Credit Default Swaption
Lemma
The price at time t [s, R] of a credit default swaption equals
C
t
= 1
{t <}
B
t
G
t
E
Q
_
G
R
B
R
A(R, U, T)((R, U, T) )
+
F
t
_
.
Dene an equivalent probability measure

Q on (, F
R
) by setting
d
Q
dQ
=
M
A
R
M
A
0
, Q-a.s.
Proposition
The price of the credit default swaption equals, for every t [s, R],
C
t
= 1
{t <}
A(t , U, T) E
Q
_
((R, U, T) )
+
F
t
_
= 1
{t <}
C
t
.
The forward CDS rate ((t , U, T), t R) is a (
Q, F)-martingale.
Brownian Case
Let the ltration F be generated by a Brownian motion W under Q.
Since M
P
and M
A
are strictly positive (Q, F)-martingales, we have that
dM
P
t
= M
P
t

P
t
dW
t
, dM
A
t
= M
A
t

A
t
dW
t
,
for some F-adapted processes
P
and
A
.
Lemma
The forward CDS rate ((t , U, T), t [0, R]) is (
Q, F)-martingale and
d(t , U, T) = (t , U, T)
t
d
W
t
where
=
P
A
and the (
Q, F)-Brownian motion

W equals
W
t
= W
t

_
t
0
A
u
du, t [0, R].
Trading Strategies
Let = (
1
,
2
) be a trading strategy, where
1
and
2
are G-adapted
processes.
The wealth of equals, for every t [s, R],
V
t
() =
1
t
S
t
() +
2
t
A(t , U, T)
and thus the pre-default wealth satises, for every t [s, R],
V
t
() =
1
t

S
t
() +
2
t

A(t , U, T).
It is enough to search for F-adapted processes
i
, i = 1, 2 such that
the equality
1
{t <}
i
t
=
i
t
holds for every t [s, R].
The next result yields a general representation for hedging strategy.
Proposition
Let the Brownian motion W be one-dimensional. The hedging strategy
= (
1
,
2
) for the credit default swaption equals, for t [s, R],

1
t
=
t
(t , U, T)
t
,
2
t
=
C
t

1
t

S
t
()
A(t , U, T)
where

is the process satisfying
C
R
A(R, U, T)
=
C
0
A(0, U, T)
+
_
R
0
t
d
W
t
.
The main issue is an explicit computation of the process

.
Market Formula
Proposition
Assume that the volatility
=
P
A
of the forward CDS spread is
deterministic. Then the pre-default value of the credit default swaption
with strike level and expiry date R equals, for every t [0, U],
C
t
=

A
t
_
t
N
_
d
+
(
t
, U t )
_
N
_
d
(
t
, U t )
_
_
where
t
= (t , U, T) and

A
t
=

A(t , U, T). Equivalently,
C
t
=

P
t
N
_
d
+
(
t
, t , R)
_
A
t
N
_
d
(
t
, t , R)
_
where

P
t
=

P(t , U, T) and
d
(
t
, t , R) =
ln(
t
/)
1
2
_
R
t
(
(u))
2
du
_
_
R
t
(
(u))
2
du
.
Assumption 1
Denition
For any u R
+
, we dene the F-martingale G
u
t
= Q( > u | F
t
) for t [0, T].
Let G
t
= G
t
t
. Then the process (G
t
, t [0, T]) is an F-supermartingale.
We also assume that G is a strictly positive process.
Assumption
There exists a family of F-adapted processes (f
x
t
; t [0, T], x R
+
) such
that, for any u R
+
,
G
u
t
=
_

u
f
x
t
dx, t [0, T].
Default Intensity
For any xed t [0, T], the random variable f
t
represents the conditional
density of with respect to the -eld F
t
, that is,
f
x
t
dx = Q( dx | F
t
).
We write f
t
t
= f
t
and we dene

t
= G
1
t
f
t
.
Lemma
Under Assumption 1, the process (M
t
, t [0, T]) given by the formula
M
t
= H
t

_
t
0
(1 H
u
)
u
du
is a G-martingale.
It can be deduced from the lemma that

= is the default intensity.
Assumption 2
Assumption
The ltration F is generated by a one-dimensional Brownian motion W.
We now work under Assumptions 1-2. We have that
For any xed u R
+
, the F-martingale G
u
satises, for t [0, T],
G
u
t
= G
u
0
+
_
t
0
g
u
s
dW
s
for some F-predictable, real-valued process (g
u
t
, t [0, T]).
For any xed x R
+
, the process (f
x
t
, t [0, T]) is an (Q, F)-martingale
and thus there exists an F-predictable process (
x
t
, t [0, T]) such that,
for t [0, T],
f
x
t
= f
x
0
+
_
t
0
x
s
dW
s
.
Survival Process
The following relationship is valid, for any u R
+
and t [0, T],
g
u
t
=
_

u
x
t
dx.
By applying the It-Wentzell-Kunita formula, we obtain the following
auxiliary result, in which we denote g
s
s
= g
s
and f
s
s
= f
s
.
Lemma
The Doob-Meyer decomposition of the survival process G equals, for every
t [0, T],
G
t
= G
0
+
_
t
0
g
s
dW
s

_
t
0
f
s
ds.
In particular, G is a continuous process.
Volatility of Pre-Default Value
Under the assumption that B, Z and A are deterministic, the volatility of
the pre-default value process can be computed explicitly in terms of
u
t
.
Recall that, for t [0, T],
f
x
t
= f
x
0
+
_
t
0
x
s
dW
s
, g
u
t
=
_

u
x
t
dx.
Corollary
If B, Z and A are deterministic then we have that, for every t [0, T],
d
S
t
=
_
(r (t ) +
t
)
S
t

t
Z(t )
_
dt + dA(t ) +
T
t
dW
t
with
T
t
= G
1
t
B(t )
T
t
where
T
t
= B
1
(T)XG
T
t
+
_
T
t
B
1
(u)Z(u)
u
t
du +
_
T
t
B
1
(u)g
u
t
dA(u).
Volatility of Forward CDS Rate
Lemma
If B, and L are deterministic then the forward CDS rate satises under

Q
d(t , U, T) = (t , U, T)
_
P
t

A
t
_
d
W
t
where the process

W, given by the formula
W
t
= W
t

_
t
0
A
u
du, t [0, R],
is a Brownian motion under

Q and
P
t
=
_
_
T
U
B
1
(u)(u)
u
t
du
__
_
T
U
B
1
(u)(u)f
u
t
du
_
1
A
t
=
_
_
Y
U
B
1
(u)g
u
t
du
__
_
T
U
B
1
(u)G
u
t
du
_
1
.
We make the following standing assumptions:
1
The default intensity process is governed by the CIR dynamics
d
t
= (
t
) dt + (
t
) dW
t
where () = a b and () = c
.
2
The default time is given by
= inf
_
t R
+
:
_
t
0
u
du
_
where is a random variable with the unit exponential distribution,
independent of the ltration F.
Model Properties
From the martingale property of f
u
we have, for every t u,
f
u
t
= E
Q
(f
u
| F
t
) = E
Q
(
u
G
u
| F
t
).
The immersion property holds between F and G so that G
t
= exp(
t
),
where
t
=
_
t
0

u
du is the hazard process. Therefore
f
s
t
= E
Q
(
s
e
s
| F
t
).
Let us denote
H
s
t
= E
Q
_
e
(
s
t
)
F
t
_
=
G
s
t
G
t
.
It is important to note that for the CIR model
H
s
t
= e
m(t ,s)n(t ,s)
t
=

H(
t
, t , s)
where

H(, t , s) is a strictly decreasing function when t < s.
Volatility of Forward CDS Rate
We assume that:
1
The tenor structure process L is deterministic.
2
The savings account is B is deterministic. We denote = B
1
.
3
We also assume that is constant.
Proposition
The volatility of the forward CDS rate satises
=
P
A
where
P
t
= (
t
)
(T)H
T
t
n(t , T) (U)H
U
t
n(t , U) +
_
T
U
r (u)(u)H
u
t
n(t , u) du
(U)H
U
t
(T)H
T
t

_
T
U
r (u)(u)H
u
t
du
and
A
t
= (
t
)
_
]U,T]
(u)H
u
t
n(t , u) dL(u)
_
]U,T]
(u)H
u
t
dL(u)
.
Equivalent Representations
One can show that
C
R
= 1
{R<}
_
_
T
U
B(R, u)
u
R
du
_
]U,T]
B(R, u)H
u
R
dL(u)
_
+
.
Straightforward computations lead to the following representation
C
R
= 1
{R<}
_
B(R, U)H
U
R

_
]U,T]
B(R, u)H
u
R
d(u)
_
+
where the function : R
+
R satises
d(u) =
ln B(R, u)
u
du + dL(u) + d1
[T,[
(u).
Auxiliary Functions
We dene auxiliary functions : R
+
R
+
and : R R
+
by setting
(x) = B(R, U)
H(x, R, U)
and
(y) =
_
]U,T]
B(R, u)
H(y, R, u) d(u).
There exists a unique F
R
-measurable random variable
R
such that
(
R
) = B(R, U)
H(
R
, R, U) =
_
]U,T]
B(R, u)
H(
R
, R, u) d(u) = (
R
).
It sufces to check that
R
=
1
((
R
)) is the unique solution to this
equation.
Explicit Valuation Formula
The payoff of the credit default swaption admits the following
representation
C
R
= 1
{R<}
_
]U,T]
B(R, u)
_
H(
R
, R, u)

H(
R
, R, u)
_
+
d(u).
Let D
0
(t , u) be the price at time t of a unit defaultable zero-coupon bond
with zero recovery maturing at u t and let B(t , u) be the price at time t
of a (default-free) unit discount bond maturing at u t .
If the interest rate process r is independent of the default intensity then
D
0
(t , u) is given by the following formula
D
0
(t , u) = 1
{t <}
B(t , u)H
u
t
.
Explicit Valuation Formula
Let P(
t
, U, u, K) stand for the price at time t of a put bond option
with strike K and expiry U written on a zero-coupon bond maturing
at u computed in the CIR model with the interest rate modeled by .
Proposition
Assume that R = U. Then the payoff of the credit default swaption equals
C
U
=
_
]U,T]
_
K(u)D
0
(U, U) D
0
(U, u)
_
+
d(u)
where K(u) = B(U, u)
H(
U
, U, u) is deterministic, since
U
=
1
().
The pre-default value of the credit default swaption equals
C
t
=
_
]U,T]
B(t , u)P(
t
, U, u,

K(u)) d(u)
where

K(u) = K(u)/B(U, u) =

H(
U
, U, u).
Hedging Strategy
1
The price P
u
t
:= P(
t
, U, u,

K(u)) of the put bond option in the CIR
model with the interest rate is known to be
P
u
t
=

K(u)H
U
t
P
U
(H
U
U

K(u) |
t
) H
u
t
P
u
(H
U
u

K(u) |
t
)
where H
u
t
=

H(
t
, t , u) is the price at time t of a zero-coupon bond
maturing at u.
2
Let us denote Z
t
= H
u
t
/H
U
t
and let us set, for every u [U, T],
P
u
(H
U
u

K(u) |
t
) =
u
(t , Z
t
).
3
Then the pricing formula for the bond put option becomes
P
u
t
=

K(u)H
U
t

U
(t , Z
t
) H
u
t

u
(t , Z
t
)
Let us recall the general representation for the hedging strategy when F is
the Brownian ltration.
Proposition
The hedging strategy = (
1
,
2
) for the credit default swaption equals, for
t [s, U],

1
t
=
t
(t , U, T)
t
,
2
t
=
C
t

1
t

S
t
()
A(t , U, T)
where

is the process satisfying
C
U
A(U, U, T)
=
C
0
A(0, U, T)
+
_
U
0
t
d
W
t
.
All terms were already computed, except for the process

.
Computation of

Recall that we are searching for the process

such that
d(
C
t
/
A(t , U, T)) =

t
d
W
t
.
Proposition
Assume that R = U. Then we have that, for every t [0, U],
t
=
1
A
t
__
]U,T]
B(t , u)
_
t
H
u
t
_
b
u
t
b
U
t
_
P
u
t
b
U
t
_
d(u)

C
t
A
t
_
where
A
t
=

A(t , U, T), H
u
t
=

H(
t
, t , u), b
u
t
= cn(t , u)
_
t
, P
u
t
= P(
t
, U, u,

K(u))
and
t
=

K(u)

U
z
(t , Z
t
)
u
(t , Z
t
) Z
t
u
z
(t , Z
t
).
Hedging Strategy
For R = U, we obtain the following nal result for hedging strategy.
Proposition
Consider the CIR default intensity model with a deterministic short-term
interest rate. The replicating strategy = (
1
,
2
) for the credit default
swaption maturing at R = U equals, for any t [0, U],

1
t
=
t
(t , U, T)
t
,
2
t
=
C
t

1
t

S
t
()
A(t , U, T)
,
where the processes
,

C and

are given in previous results.
Note that for R < U the problem remains open, since a closed-form solution
for the process

is not readily available in this case.
Credit Default Index Swap
Credit Default Index Swaption
Loss-Adjusted Forward CDIS
1
A credit default index swap (CDIS) is a standardized contract that is
based upon a xed portfolio of reference entities.
2
At its conception, the CDIS is referenced to n xed companies that are
chosen by market makers.
3
The reference entities are specied to have equal weights.
4
If we assume each has a nominal value of one then, because of the
equal weighting, the total notional would be n.
5
By contrast to a standard single-name CDS, the buyer of the CDIS
provides protection to the market makers.
6
By purchasing a CDIS from market makers the investor is not receiving
protection, rather they are providing it to the market makers.
1
In exchange for the protection the investor is providing, the market
makers pay the investor a periodic xed premium, otherwise known as
the credit default index spread.
2
The recovery rate [0, 1] is predetermined and identical for all
reference entities in the index.
3
By purchasing the index the investor is agreeing to pay the market
makers 1 for any default that occurs before maturity.
4
Following this, the nominal value of the CDIS is reduced by one; there is
no replacement of the defaulted rm.
5
This process repeats after every default and the CDIS continues on until
maturity.
Default Times and Filtrations
1
Let
1
, . . . ,
n
represent default times of reference entities.
2
We introduce the sequence
(1)
< <
(n)
of ordered default times
associated with
1
, . . . ,
n
. For brevity, we write =
(n)
.
3
We thus have G = H
(n)

F, where H
(n)
is the ltration generated by
the indicator process H
(n)
t
= 1
{ t }
of the last default and the ltration

F
equals

F = F H
(1)
H
(n1)
.
4
We are interested in events of the form { t } and { > t } for a xed t .
5
Morini and Brigo (2007) refer to these events as the armageddon and
the no-armageddon events. We use instead the terms collapse event
and the pre-collapse event.
6
The event { t } corresponds to the total collapse of the reference
portfolio, in the sense that all underlying credit names default either prior
to or at time t .
Basic Lemma
1
We set

F
t
= Q( t |
F
t
) for every t R
+
.
2
Let us denote by

G
t
= 1

F
t
= Q( > t |
F
t
) the corresponding survival
process with respect to the ltration

F and let us temporarily assume that
the inequality

G
t
> 0 holds for every t R
+
.
3
Then for any Q-integrable and

F
T
-measurable random variable Y we
have that
E
Q
(1
{T< }
Y | G
t
) = 1
{t < }

G
1
t
E
Q
(
G
T
Y |
F
t
).
Lemma
Assume that Y is some G-adapted stochastic process. Then there exists a
unique

F-adapted process

Y such that, for every t [0, T],
Y
t
= 1
{t < }
Y
t
.
The process

Y is termed the pre-collapse value of the process Y.
Notation and Assumptions
We write T
0
= T < T
1
< < T
J
to denote the tenor structure of the
forward-start CDIS, where:
1
T
0
= T is the inception date;
2
T
J
is the maturity date;
3
T
j
is the j th fee payment date for j = 1, 2, . . . , J;
4
a
j
= T
j
T
j 1
for every j = 1, 2, . . . , J.
The process B is an F-adapted (or, at least,

F-adapted) and strictly positive
process representing the price of the savings account.
The underlying probability measure Q is interpreted as a martingale measure
associated with the choice of B as the numeraire asset.
Forward Credit Default Index Swap
Denition
The discounted cash ows for the seller of the forward CDIS issued at time
s [0, T] with an F
s
-measurable spread are, for every t [s, T],
D
n
t
= P
n
t
A
n
t
,
where
P
n
t
= (1 )B
t
n
i =1
B
1
i
1
{T<
i
T
J
}
A
n
t
= B
t
J
j =1
a
j
B
1
T
j
n
i =1
_
1 1
{T
j
i
}
_
are discounted payoffs of the protection leg and the fee leg per one basis
point, respectively. The fair price at time t [s, T] of a forward CDIS equals
S
n
t
() = E
Q
(D
n
t
| G
t
) = E
Q
(P
n
t
| G
t
) E
Q
(A
n
t
| G
t
).
Forward Credit Default Index Swap
1
The quantities P
n
t
and A
n
t
are well dened for any t [0, T] and they do
not depend on the issuance date s of the forward CDIS under
consideration.
2
They satisfy
P
n
t
= 1
{T< }
P
n
t
, A
n
t
= 1
{T< }
A
n
t
.
3
For brevity, we will write J
t
to denote the reduced nominal at time
t [s, T], as given by the formula
J
t
=
n
i =1
_
1 1
{t
i
}
_
.
4
In what follows, we only require that the inequality

G
t
> 0 holds for every
t [s, T
1
], so that, in particular,

G
T
1
= Q( > T
1
|
F
T
1
) > 0.
Pre-collapse Price
Lemma
The price at time t [s, T] of the forward CDIS satises
S
n
t
() = 1
{t < }
G
1
t
E
Q
(D
n
t
|
F
t
) = 1
{t < }
S
n
t
(),
where the pre-collapse price of the forward CDIS satises

S
n
t
() =

P
n
t

A
n
t
,
where
P
n
t
=

G
1
t
E
Q
(P
n
t
|
F
t
) = (1 )
G
1
t
B
t
E
Q
_
n
i =1
B
1
i
1
{T<
i
T
J
}
F
t
_
A
n
t
=

G
1
t
E
Q
(A
n
t
|
F
t
) =

G
1
t
B
t
E
Q
_
J
j =1
a
j
B
1
T
j
J
T
j
F
t
_
.
The process

A
n
t
may be thought of as the pre-collapse PV of receiving risky
one basis point on the forward CDIS payment dates T
j
on the residual
nominal value J
T
j
. The process

P
n
t
represents the pre-collapse PV of the
protection leg.
Pre-Collapse Fair CDIS Spread
Since the forward CDIS is terminated at the moment of the nth default with no
further payments, the forward CDS spread is dened only prior to .
Denition
The pre-collapse fair forward CDIS spread is the

F
t
-measurable random
variable
n
t
such that

S
n
t
(
n
t
) = 0.
Lemma
Assume that

G
T
1
= Q( > T
1
|
F
T
1
) > 0. Then the pre-collapse fair forward
CDIS spread satises, for t [0, T],
n
t
=
P
n
t
A
n
t
=
(1 ) E
Q
_
n
i =1
B
1
i
1
{T<
i
T
J
}
F
t
_
E
Q
_
J
j =1
a
j
B
1
T
j
J
T
j
F
t
_ .
The price of the forward CDIS admits the following representation
S
n
t
() = 1
{t < }
A
n
t
(
n
t
).
Market Convention for Valuing a CDIS
Market quote for the quantity

A
n
t
, which is essential in marking-to-market of a
CDIS, is not directly available. The market convention for approximation of
the value of

A
n
t
hinges on the following postulates:
1
all rms are identical from time t onwards (homogeneous portfolio);
therefore, we just deal with a single-name case, so that either all rms
default or none;
2
the implied risk-neutral default probabilities are computed using a at
single-name CDS curve with a constant spread equal to
n
t
.
Then
A
n
t
J
t
PV
t
(
n
t
),
where PV
t
(
t
) is the risky present value of receiving one basis point at all
CDIS payment dates calibrated to a at CDS curve with spread equal to
n
t
,
where
n
t
is the quoted CDIS spread at time t .
The conventional market formula for the value of the CDIS with a xed spread
reads, on the pre-collapse event {t < },
S
t
() = J
t
PV
t
(
n
t
)(
n
t
).
Market Payoff of a Credit Default Index Swaption
1
The conventional market formula for the payoff at maturity U T of the
payer credit default index swaption with strike level reads
C
U
=
_
1
{U< }
PV
U
_
n
U
_
J
U
(
n
U

n
0
) 1
{U< }
PV
U
()n(
n
0
) + L
U
_
+
,
where L stands for the loss process for our portfolio so that, for every
t R
+
,
L
t
= (1 )
n
i =1
1
{
i
t }
.
2
The market convention is due to the fact that the swaption has physical
settlement and the CDIS with spread is not traded. If the swaption is
exercised, its holder takes a long position in the on-the-run index and is
compensated for the difference between the value of the on-the-run
index and the value of the (non-traded) index with spread , as well as
for defaults that occurred in the interval [0, U].
Put-Call Parity for Credit Default Index Swaptions
1
For the sake of brevity, let us denote, for any xed > 0,
f (, L
U
) = L
U
1
{U< }
PV
U
()n(
n
0
).
2
Then the payoff of the payer credit default index swaption entered at time
0 and maturing at U equals
C
U
=
_
1
{U< }
PV
U
_
n
U
_
J
U
(
n
U

n
0
) + f (, L
U
)
_
+
,
whereas the payoff of the corresponding receiver credit default index
swaption satises
P
U
=
_
1
{U< }
PV
U
_
n
U
_
J
U
(
n
0

n
U
) f (, L
U
)
_
+
.
3
This leads to the following equality, which holds at maturity date U
C
U
P
U
= 1
{U< }
PV
U
_
n
U
_
J
U
(
n
U

n
0
) + f (, L
U
).
Model Payoff of a Credit Default Index Swaption
1
The model payoff of the payer credit default index swaption entered at
time 0 with maturity date U and strike level equals
C
U
= (S
n
U
() + L
U
)
+
or, more explicitly
C
U
=
_
1
{U< }
A
n
U
(
U
) + L
U
_
+
.
2
To formally derive obtain the model payoff from the market payoff, it
sufces to postulate that
PV
U
()n PV
U
_
U
_
J
U

A
n
U
.
1
Since L
U
0 and
L
U
= 1
{U< }
L
U
+1
{U }
L
U
the payoff C
U
can also be represented as follows
C
U
= (S
n
U
() +1
{U< }
L
U
)
+
+1
{U }
L
U
= (S
a
U
())
+
+ C
L
U
,
where we denote
S
a
U
() = S
n
U
() +1
{U< }
L
U
and
C
L
U
= 1
{U }
L
U
.
2
The quantity S
a
U
() represents the payoff at time U of the loss-adjusted
forward CDIS.
1
The discounted cash ows for the seller of the loss-adjusted forward
CDIS (that is, for the buyer of the protection) are, for every t [0, U],
D
a
t
= P
a
t
A
n
t
,
where
P
a
t
= P
n
t
+ B
t
B
1
U
1
{U< }
L
U
.
2
It is essential to observe that the payoff D
a
U
is the U-survival claim, in the
sense that
D
a
U
= 1
{U< }
D
a
U
.
3
Any other adjustments to the payoff P
n
t
or A
n
t
are also admissible,
provided that the properties
P
a
U
= 1
{U< }
P
a
U
, A
a
U
= 1
{U< }
A
a
U
hold.
Price of the Loss-Adjusted Forward CDIS
Lemma
The price of the loss-adjusted forward CDIS equals, for every t [0, U],
S
a
t
() = 1
{t < }
G
1
t
E
Q
(D
a
t
|
F
t
) = 1
{t < }
S
a
t
(),
where the pre-collapse price satises

S
a
t
() =

P
a
t

A
n
t
, where in turn
P
a
t
=

G
1
t
E
Q
(P
a
t
|
F
t
),

A
n
t
=

G
1
t
E
Q
(A
n
t
|
F
t
)
or, more explicitly,
P
a
t
=

G
1
t
B
t
E
Q
_
(1 )
n
i =1
B
1
i
1
{T<
i
T
J
}
+1
{U< }
B
1
U
L
U
F
t
_
and
A
n
t
=

G
1
t
B
t
E
Q
_
J
j =1
a
j
B
1
T
j
J
T
j
F
t
_
.
Pre-Collapse Loss-Adjusted Fair CDIS Spread
We are in a position to dene the fair loss-adjusted forward CDIS spread.
Denition
The pre-collapse loss-adjusted fair forward CDIS spread at time t [0, U] is
the

F
t
-measurable random variable
a
t
such that

S
a
t
(
a
t
) = 0.
Lemma
Assume that

G
T
1
= Q( > T
1
|
F
T
1
) > 0. Then the pre-collapse loss-adjusted
fair forward CDIS spread satises, for t [0, U],
a
t
=
P
a
t
A
n
t
=
E
Q
_
(1 )
n
i =1
B
1
i
1
{T<
i
T
J
}
+1
{U< }
B
1
U
L
U
F
t
_
E
Q
_
J
j =1
a
j
B
1
T
j
J
T
j
F
t
_ .
The price of the forward CDIS has the following representation, for t [0, T],
S
a
t
() = 1
{t < }
A
n
t
(
a
t
).
Model Pricing of Credit Default Index Swaptions
1
It is easy to check that the model payoff can be represented as follows
C
U
= 1
{U< }
A
n
U
(
a
U
)
+
+1
{U }
L
U
.
2
The price at time t [0, U] of the credit default index swaption is thus
given by the risk-neutral valuation formula
C
t
= B
t
E
Q
_
1
{U< }
B
1
U

A
n
U
(
a
U
)
+
G
t
_
+ B
t
E
Q
_
1
{U }
B
1
U
L
U
G
t
_
.
3
Using the ltration

F, we can obtain a more explicit representation for the
rst term in the formula above, as the following result shows.
Lemma
The price at time t [0, U] of the payer credit default index swaption equals
C
t
= E
Q
_
G
U
B
1
U

A
n
U
(
a
U
)
+
F
t
_
+ B
t
E
Q
_
1
{U }
B
1
U
L
U
G
t
_
.
1
The random variable Y = B
1
U

A
n
U
(
a
U
)
+
is manifestly

F
U
-measurable
and Y = 1
{U< }
Y. Hence the equality is an immediate consequence of
the basic lemma.
2
On the collapse event {t } we have 1
{U }
B
1
U
L
U
= B
1
U
n(1 )
and thus the pricing formula reduces to
C
t
= B
t
E
Q
_
1
{U }
B
1
U
L
U
G
t
_
= n(1)E
Q
_
B
1
U
G
t
_
= n(1)B(t , T),
where B(t , T) is the price at t of the U-maturity risk-free zero-coupon
bond.
1
Let us thus concentrate on the pre-collapse event {t < }. We now have
C
t
= C
a
t
+ C
L
t
, where
C
a
t
= B
t
G
1
t
E
Q
_
G
U
B
1
U

A
n
U
(
a
U
)
+
F
t
_
and
C
L
t
= B
t
E
Q
_
1
{U >t }
B
1
U
L
U

F
t
_
.
The last equality follows from the well known fact that on {t < } any
G
t
-measurable event can be represented by an

F
t
-measurable event, in
the sense that for any event A G
t
there exists an event

A

F
t
such
that 1
{t < }
A = 1
{t < }
A.
1
The computation of C
L
t
relies on the knowledge of the risk-neutral
conditional distribution of given

F
t
and the term structure of interest
rates, since on the event {U > t } we have B
1
U
L
U
= B
1
U
n(1 ).
2
For C
a
t
, we dene an equivalent probability measure

Q on (,
F
U
)
d
Q
dQ
= c
G
U
B
1
U

A
n
U
, Q-a.s.
3
Note that the process
t
= c
G
t
B
1
t

A
n
t
, t [0, U], is a strictly positive
F-martingale under Q, since

t
= c
G
t
B
1
t

A
n
t
= c E
Q
_
J
j =1
a
j
B
1
T
j
J
T
j
F
t
_
and Q( > T
j
|
F
T
j
) =

G
T
j
> 0 for every j .
4
Therefore, for every t [0, U],
d
Q
dQ
F
t
= E
Q
(
U
|
F
t
) =
t
, Q-a.s.
Model Pricing Formula for Credit Default Index Swaptions
Lemma
The price at time t [0, U] of the payer credit default index swaption on the
pre-collapse event {t < } equals
C
t
=

A
n
t
E
Q
_
(
a
U
)
+

F
t
_
+ B
t
E
Q
_
1
{U >t }
B
1
U
L
U

F
t
_
.
The next lemma establishes the martingale property of the process
a
under

Q.
Lemma
The pre-collapse loss-adjusted fair forward CDIS spread
a
t
, t [0, U], is a
strictly positive

F-martingale under

Q.
Black Formula for Credit Default Index Swaptions
1
Our next goal is to establish a suitable version of the Black formula for
the credit default index swaption.
2
To this end, we postulate that the pre-collapse loss-adjusted fair forward
CDIS spread satises
a
t
=
a
0
+
_
t
0
a
u
d
W
u
, t [0, U],
where

W is the one-dimensional standard Brownian motion under

Q with
respect to

F and is an

F-predictable process.
3
The assumption that the ltration

F is the Brownian ltration would be too
restrictive, since

F = F H
(1)
H
(n1)
and thus

F will typically need
to support also discontinuous martingales.
Market Pricing Formula for Credit Default Index Swaptions
Proposition
Assume that the volatility of the pre-collapse loss-adjusted fair forward
CDIS spread is a positive function. Then the pre-default price of the payer
credit default index swaption equals, for every t [0, U] on the pre-collapse
event {t < },
C
t
=

A
n
t
_
a
t
N
_
d
+
(
a
t
, t , U)
_
N
_
d
(
a
t
, t , U)
_
_
+ C
L
t
or, equivalently,
C
t
=

P
a
t
N
_
d
+
(
a
t
, t , U)
_
A
n
t
N
_
d
(
a
t
, t , U)
_
+ C
L
t
,
where
d
(
a
t
, t , U) =
ln(
a
t
/)
1
2
_
U
t

2
(u) du
_ _
U
t

2
(u) du
_
1/2
.
Approximation
Proposition
The price of a payer credit default index swaption can be approximated as
follows
C
t
1
{t < }
A
n
t
_
n
t
N
_
d
+
(
n
t
, t , U)
_
(
L
t
)N
_
d
(
n
t
, t , U)
_
_
,
where for every t [0, U]
d
(
n
t
, t , U) =
ln(
n
t
/(
L
t
))
1
2
_
U
t

2
(u) du
_ _
U
t

2
(u) du
_
1/2
and
L
t
= E
Q
_
(A
n
U
)
1
L
U
|
F
t
_
.
Comments
1
Under usual circumstances, the probability of all defaults occurring prior
to U is expected to be very low.
2
However, as argued by Morini and Brigo (2007), this assumption is not
always justied, in particular, it is not suitable for periods when the
market conditions deteriorate.
3
It is also worth mentioning that since we deal here with the risk-neutral
probability measure, the probabilities of default events are known to
drastically exceed statistically observed default probabilities, that is,
probabilities of default events under the physical probability measure.
One-Period Case
One- and Two-Period Case
Towards Generic Swap Models
Conclusions
One-Period Case
Conclusions
Notation
1
Let (, G, F, Q) be a ltered probability space, where F = (F
t
)
t [0,T]
is a
ltration such that F
0
is trivial.
2
We assume that the random time dened on this space is such that
the F-survival process G
t
= Q( > t | F
t
) is positive.
3
The probability measure Q is interpreted as the risk-neutral measure.
4
Let 0 < T
0
< T
1
< < T
n
be a xed tenor structure and let us write
a
i
= T
i
T
i 1
.
5
We denote

a
i
= a
i
/(1
i
) where
i
is the recovery rate if default occurs
between T
i 1
and T
i
.
6
We denote by (t , T) the default-free discount factor over the time
period [t , T].
One-Period Case
Conclusions
Bottom-up Approach under Deterministic Interest Rates
1
Assume rst that the interest rate is deterministic.
2
The pre-default forward CDS spread
i
corresponding to the
single-period forward CDS starting at time T
i 1
and maturing at T
i
equals
1 +
a
i
i
t
=
E
Q
_
(t , T
i
)1
{>T
i 1
}
F
t
_
E
Q
_
(t , T
i
)1
{>T
i
}
F
t
_ , t [0, T
i 1
].
3
Since the interest rate is deterministic, we obtain, for i = 1, . . . , n,
1 +
a
i
i
t
=
Q( > T
i 1
| F
t
)
Q( > T
i
| F
t
)
, t [0, T
i 1
],
and thus
Q( > T
i
| F
t
)
Q( > T
0
| F
t
)
=
i
j =1
1
1 +
a
j
j
t
, t [0, T
0
].
One-Period Case
Conclusions
Auxiliary Probability Measure P
We dene the probability measure P equivalent to Q on (, F
T
) by setting, for
every t [0, T],
t
=
dP
dQ
F
t
=
Q( > T
n
| F
t
)
Q( > T
n
| F
0
)
.
Lemma
For every i = 1, . . . , n, the process Z
,i
given by
Z
,i
t
=
n
j =i +1
_
1 +
a
j
j
t
_
, t [0, T
i
],
is a positive (P, F)-martingale.
One-Period Case
Conclusions
CDS Martingale Measures
1
For any i = 1, . . . , n we dene the probability measure P
i
equivalent to P
on (, F
T
) by setting (note that Z
,n
t
= 1 and thus P
n
= P)
dP
i
dP
F
t
= c
i
Z
,i
t
=
Q( > T
i
)
Q( > T
n
)
n
j =i +1
_
1 +
a
j
j
t
_
.
2
Assume that the PRP holds under P = P
n
with the R
k
-valued spanning
(P, F)-martingale M. Then the PRP is also valid with respect to F under
any probability measure P
i
for i = 1, . . . , n.
3
The positive process
i
is a (P
i
, F)-martingale and thus it satises, for
i = 1, . . . , n,
i
t
=
i
0
+
_
(0,t ]
i
s
i
s
d
i
(M)
s
for some R
k
-valued, F-predictable process
i
, where
i
(M) is the
P
i
-Girsanov transform of M
i
(M)
t
= M
i
t

_
(0,t ]
(Z
i
s
)
1
d[Z
i
, M]
s
.
One-Period Case
Conclusions
Dynamics of Forward CDS Spreads
Proposition
Let the processes
i
, i = 1, . . . , n, be dened by
1 +
a
i
i
t
=
E
Q
_
(t , T
i
)1
{>T
i 1
}
F
t
_
E
Q
_
(t , T
i
)1
{>T
i
}
F
t
_ , t [0, T
i 1
].
Assume that the PRP holds with respect to F under P with the spanning
(P, F)-martingale M = (M
1
, . . . , M
k
). Then there exist R
k
-valued,
F-predictable processes
i
such that the joint dynamics of processes
i
, i = 1, . . . , n under P are given by
d
i
t
=
k
l =1
i
t
i ,l
t
dM
l
t

n
j =i +1
a
j
i
t
j
t
1 +
a
j
j
t
k
l ,m=1
i ,l
t

j ,m
t
d[M
l ,c
, M
m,c
]
t
1
Z
i
t
Z
i
t
k
l =1
i
t
i ,l
t
M
l
t
.
One-Period Case
Conclusions
Top-down Approach: First Step
Proposition
Assume that:
(i) the positive processes
i
, i = 1, . . . , n, are such that the processes
Z
,i
, i = 1, . . . , n are (P, F)-martingales, where
Z
,i
t
=
n
j =i +1
_
1 +
a
j
j
t
_
.
(ii) M = (M
1
, . . . , M
k
) is a spanning (P, F)-martingale.
(iii)
i
, i = 1, . . . , n are R
k
-valued, F-predictable processes.
Then:
(i) for every i = 1, . . . , n, the process
i
is a (P
i
, F)-martingale where
dP
i
dP
F
t
= c
i
n
j =i +1
_
1 +
a
j
j
t
_
,
(ii) the joint dynamics of processes
i
, i = 1, . . . , n under P are given by the
previous proposition.
One-Period Case
Conclusions
Top-down Approach: Second Step
1
We will now construct a default time consistent with the dynamics of
forward CDS spreads. Let us set
M
i 1
T
i 1
=
i 1
j =1
1
1 +
a
j
j
T
i 1
, M
i
T
i
=
i
j =1
1
1 +
a
j
j
T
i
.
2
Since the process

a
i
i
is positive, we obtain, for every i = 0, . . . , n,
G
T
i
:= M
i
T
i
=
M
i 1
T
i 1
1 +
a
i
i
T
i
M
i 1
T
i 1
=: G
i 1
T
i 1
.
3
The process G
T
i
= M
i
T
i
is thus decreasing for i = 0, . . . , n.
4
We make use of the canonical construction of default time taking
values in {T
0
, . . . , T
n
}.
5
We obtain, for every i = 0, . . . , n,
P( > T
i
| F
T
i
) = G
T
i
=
i
j =1
1
1 +
a
j
j
T
i
.
One-Period Case
Conclusions
Bottom-up Approach under Independence
Assume that we are given a model for Libors (L
1
, . . . , L
n
) where
L
i
= L(t , T
i 1
) and CDS spreads (
1
, . . . ,
n
) in which:
1
The default intensity generates the ltration F
.
2
The interest rate process r generates the ltration F
r
.
3
The probability measure Q is the spot martingale measure.
4
The H-hypothesis holds, that is, F
Q
G, where F = F
r
F
.
5
The PRP holds with the (Q, F)-spanning martingale M.
Lemma
It is possible to determine the joint dynamics of Libors and CDS spreads
(L
1
, . . . , L
n
,
1
, . . . ,
n
) under any martingale measure P
i
.
One-Period Case
Conclusions
Top-down Approach under Independence
To construct a model we assume that:
1
A martingale M = (M
1
, . . . , M
k
) has the PRP with respect to (P, F).
2
The family of process Z
i
given by
Z
L,,i
t
:=
n
j =i +1
(1 + a
j
L
j
t
)(1 +
a
j
j
t
)
are martingales on the ltered probability space (, F, P).
3
Hence there exists a family of probability measures P
i
, i = 1, . . . , n
on (, F
T
) with the densities
dP
i
dP
= c
i
Z
L,,i
.
One-Period Case
Conclusions
Dynamics of LIBORs and CDS Spreads
Proposition
The dynamics of L
i
and
i
under P
n
with respect to the spanning
(P, F)-martingale M are given by
dL
i
t
=
k
l =1
i ,l
t
dM
l
t

n
j =i +1
a
j
1 + a
j
L
j
t
k
l ,m=1
i ,l
t

j ,m
t
d[M
l ,c
, M
m,c
]
t
j =i +1
a
j
1 +
a
j
j
t
k
l ,m=1
i ,l
t

j ,m
t
d[M
l ,c
, M
m,c
]
t

1
Z
i
t
Z
i
t
k
l =1
i ,l
t
M
l
t
and
d
i
t
=
k
l =1
i ,l
t
dM
l
t

n
j =i +1
a
j
1 + a
j
L
j
t
k
l ,m=1
i ,l
t

j ,m
t
d[M
l ,c
, M
m,c
]
t
j =i +1
a
j
1 +
a
j
j
t
k
l ,m=1
i ,l
t

j ,m
t
d[M
l ,c
, M
m,c
]
t

1
Z
i
t
Z
i
t
k
l =1
i ,l
t
M
l
t
.
One-Period Case
Conclusions
Bottom-up Approach: One- and Two-Period Spreads
1
Let (, G, F, Q) be a ltered probability space, where F = (F
t
)
t [0,T]
is a
ltration such that F
0
is trivial.
2
We assume that the random time dened on this space is such that
the F-survival process G
t
= Q( > t | F
t
) is positive.
3
The probability measure Q is interpreted as the risk-neutral measure.
4
Let 0 < T
0
< T
1
< < T
n
be a xed tenor structure and let us write
a
i
= T
i
T
i 1
and

a
i
= a
i
/(1
i
)
5
We no longer assume that the interest rate is deterministic.
6
We denote by (t , T) the default-free discount factor over the time
period [t , T].
One-Period Case
Conclusions
One-Period CDS Spreads
The one-period forward CDS spread
i
=
i 1,i
satises, for t [0, T
i 1
],
1 +
a
i
i
t
=
E
Q
_
(t , T
i
)1
{>T
i 1
}
F
t
_
E
Q
_
(t , T
i
)1
{>T
i
}
F
t
_ .
Let A
i 1,i
be the one-period CDS annuity
A
i 1,i
t
=

a
i
E
Q
_
(t , T
i
)1
{>T
i
}
F
t
_
and let
P
i 1,i
t
= E
Q
_
(t , T
i
)1
{>T
i 1
}
F
t
_
E
Q
_
(t , T
i
)1
{>T
i
}
F
t
_
.
Then
i
t
=
P
i 1,i
t
A
i 1,i
t
, t [0, T
i 1
].
One-Period Case
Conclusions
One-Period CDS Spreads
Let A
i 2,i
stand for the two-period CDS annuity
A
i 2,i
t
=

a
i 1
E
Q
_
(t , T
i 1
)1
{>T
i 1
}
F
t
_
+
a
i
E
Q
_
(t , T
i
)1
{>T
i
}
F
t
_
and let
P
i 2,i
t
=
i
j =i 1
_
E
Q
_
(t , T
j
)1
{>T
j 1
}
F
t
_
E
Q
_
(t , T
j
)1
{>T
j
}
F
t
_ _
.
The two-period CDS spread
i
=
i 2,i
is given by the following expression

i
t
=
i 2,i
t
=
P
i 2,i
t
A
i 2,i
t
=
P
i 2,i 1
t
+ P
i 1,i
t
A
i 2,i 1
t
+ A
i 1,i
t
, t [0, T
i 1
].
One-Period Case
Conclusions
One-Period CDS Measures
1
Our aim is to derive the semimartingale decomposition of
i
, i = 1, . . . , n
and
i
, i = 2, . . . , n under a common probability measure.
2
We start by noting that the process A
n1,n
is a positive (Q, F)-martingale
and thus it denes the probability measure P
n
on (, F
T
).
3
The following processes are easily seen to be (P
n
, F)-martingales
A
i 1,i
t
A
n1,n
t
=
n
j =i +1
a
j
(
j
t

j
t
)
a
j 1
(
j 1
t

j
t
)
=
a
n
a
i
n
j =i +1

j
t

j
t
j 1
t

j
t
.
4
Given this family of positive (P
n
, F)-martingales, we dene a family of
probability measures P
i
for i = 1, . . . , n such that
i
is a martingale
under P
i
.
One-Period Case
Conclusions
Two-Period CDS Measures
1
For every i = 2, . . . , n, the following process is a (P
i
, F)-martingale
A
i 2,i
t
A
i 1,i
t
=
a
i 1
E
Q
_
(t , T
i 1
)1
{>T
i 1
}
F
t
_
+
a
i
E
Q
_
(t , T
i
)1
{>T
i
}
F
t
_
E
Q
_
(t , T
i
)1
{>T
i
}
F
t
_
=

a
i 1
_
A
i 2,i 1
t
A
i 1,i
t
+ 1
_
=

a
i
_

i
t

i
t
i 1
t

i
t
+ 1
_
.
2
Therefore, we can dene a family of the associated probability measures
P
i
on (, F
T
), for every i = 2, . . . , n.
3
It is obvious that
i
is a martingale under

P
i
for every i = 2, . . . , n.
One-Period Case
Conclusions
One and Two-Period CDS Measures
We will summarise the above in the following diagram
Q
dP
n
dQ
P
n
dP
n1
dP
n
P
n1
dP
n2
dP
n1
. . . P
2
P
1
d
P
n
dP
n
_
d
P
n1
dP
n1
_
d
P
2
dP
2
P
n

P
n1
. . .

P
2
where
dP
n
dQ
= A
n1,n
t
dP
i
dP
i +1
=
A
i 1,i
t
A
i ,i +1
t
=
a
i +1
a
i
_

i +1
t

i +1
t
i
t

i +1
t
_
d
P
i
dP
i
=
A
i 2,i
t
A
i 1,i
t
=

a
i
_

i
t

i
t
i 1
t

i
t
+ 1
_
.
One-Period Case
Conclusions
Bottom-up Approach: Joint Dynamics
1
We are in a position to calculate the semimartingale decomposition of
(
1
, . . . ,
n
,
2
, . . . ,
n
) under P
n
.
2
It sufces to use the following Radon-Nikodm densities
dP
i
dP
n
=
A
i 1,i
t
A
n1,n
t
=
a
n
a
i
n
j =i +1

j
t

j
t
j 1
t

j
t
d
P
i
dP
n
=
A
i 2,i
t
A
n1,n
t
=

a
n
_

i
t

i
t
i 1
t

i
t
+ 1
_
n
j =i +1

j
t

j
t
j 1
t

j
t
=

a
n
_
_
n
j =i

j
t

j
t
j 1
t

j
t
+
n
j =i +1

j
t

j
t
j 1
t

j
t
_
_
=

a
i 1
dP
i 1
dP
n
+
a
i
dP
i
dP
n
.
3
Explicit formulae for the joint dynamics of one and two-period spreads
are available.
One-Period Case
Conclusions
Top-down Approach: Postulates
1
The processes
1
, . . . ,
n
and
2
, . . . ,
n
are F-adapted.
2
,i
Z
,i
t
=
c
n
c
i
n
j =i +1

j
t

j
t
j 1
t

j
t
is a positive (P, F)-martingale where c
1
, . . . , c
n
are constants.
3
,i
given by the formula
Z
,i
=

c
i
(Z
,i
+ Z
,i 1
) =

c
i
i 1
i 1

i
Z
,i
is a positive (P, F)-martingale where

c
2
, . . . ,
c
n
are constants.
4
The process M = (M
1
, . . . , M
k
) is the (P, F)-spanning martingale.
5
Probability measures P
i
and

P
i
have the density processes Z
,i
and Z
,i
.
In particular, the equality P
n
= P holds, since Z
,n
= 1.
6
Processes
i
and
i
are martingales under P
i
and

P
i
, respectively.
One-Period Case
Conclusions
Top-down Approach: Lemma
Lemma
Let M = (M
1
, . . . , M
k
) be the (P, F)-spanning martingale. For any
i = 1, . . . , n, the process X
i
admits the integral representation
i
t
=
_
(0,t ]
i
s
d
i
(M)
s
and

i
t
=
_
(0,t ]
i
s
d
i
(M)
s
where
i
= (
i ,1
, . . . ,
i ,k
) and
i
= (
i ,1
, . . . ,
i ,k
) are R
k
-valued,
F-predictable processes that can be chosen arbitrarily. The (P
i
, F)-martingale
i
(M
l
) is given by
i
(M
l
)
t
= M
l
t

_
(ln Z
,i
)
c
, M
l ,c
_
t
0<st
1
Z
,i
s
Z
,i
s
M
l
s
.
An analogous formula holds for the Girsanov transform

i
(M
l
).
One-Period Case
Conclusions
Top-down Approach: Joint Dynamics
Proposition
The semimartingale decomposition of the (P
i
, F)-spanning martingale
i
(M)
under the probability measure P
n
= P is given by, for i = 1, . . . , n,
i
(M)
t
= M
t

n
j =i +1
_
(0,t ]
(
j 1
s

j
s
)
j
s
d[M
c
]
s
(
j
s

j
s
)(
j 1
s

j
s
)
j =i +1
_
(0,t ]
j
s
d[M
c
]
s

j
s

j
s
j =i +1
_
(0,t ]
j 1
s
d[M
c
]
s
j 1
s

j
s
0<st
1
Z
,i
s
Z
,i
s
M
s
.
An analogous formula holds for

i
(M). Hence the joint dynamics of the
process (
1
, . . . ,
n
,
2
, . . . ,
n
) under P = P
n
are explicitly known.
One-Period Case
Conclusions
Let (, F, P) be a ltered probability space. Suppose that we are given a
family of swaps S = {
1
, . . . ,
l
} and a family of processes {Z
1
, . . . , Z
l
}
satisfying the following conditions for every j = 1, . . . , l :
1
the process
j
is a positive special semimartingale,
2
the process
j
Z
j
is a (P, F)-martingale,
3
the process Z
j
is a positive (P, F)-martingale with Z
j
0
= 1,
4
the process Z
j
is uniquely expressed as a function of some subset of
swaps in S, specically, Z
j
= f
j
(
n
1
, . . . ,
n
k
) where f
j
: R
k
R is a C
2
function in variables belonging to {
n
1
, . . . ,
n
k
} S.
One-Period Case
Conclusions
Volatility-Based Modelling
1
For the purpose of modelling, we select a (P, F)-martingale M and we
dene
j
under P
j
as follows
j
t
=
_
t
0
j
s
j
s
d
j
(M)
s
.
2
Therefore, specifying
j
is equivalent to specifying the volatility
j
.
3
The martingale part of
j
can be expressed as
(
j
)
m
t
=
_
t
0
j
s
j
s
d
j
(M)
s
_
(0,t ]
Z
j
s
j
s
j
s
d
_
1
Z
j
,
j
(M)
_
s
=
_
t
0
j
s
j
s
dM
j
s
where M
j
is a (P, F)-martingale.
4
The Radon-Nikodm density process Z
j
has the following decomposition
Z
j
t
=
k
i =1
_
[0,t )
f
j
x
i
(
n
1
s
, . . . ,
n
k
s
)
n
i
s

n
i
s
dM
n
i
s
.
5
Hence the choice of volatilities completely species the model.

Valuation of Credit Default Swaptions and Credit Default Index Swaptions Nice - Rutkowski

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Valuation of Credit Default Swaptions and Credit Default Index Swaptions Nice - Rutkowski

Încărcat de

Drepturi de autor:

Formate disponibile

Credit Default Swaptions

Credit Default Index Swaptions

P(t , U, T) is the pre-default value of the protection leg,

Recall that we are searching for the process

F-martingale under Q, since

S-ar putea să vă placă și