7 - Default Swaptions, Index Options, and CMDS

Default swaptions
Index options
Constant maturity default swaps (CMDS)
Interest and Credit Models

7. Default Swaptions, Index Options, and CMDS
Andrew Lesniewski
Baruch College
New York
Spring 2019
A. Lesniewski Interest and Credit Models

Default swaptions
Index options
Outline
1 Default swaptions
2 Index options
3 Constant maturity default swaps (CMDS)

Default swaptions
Index options
Default swaption
A default swaption or CDS option is an OTC instrument which grants the holder
the right, but not obligation, to enter into a (single name) CDS on a contractually
specified date T0 and at a contractually specified spread C.
An option to buy protection is called a payer swaption, while an option to sell
protection is called a receiver swaption.
It is common (especially outside of the USD market) that default swaptions have
a knock out feature built in.
Namely, if the reference name underlying the CDS defaults prior to T0 , the
swaption is knocked out and the contract is nullified.

Default swaptions
Index options
Default swaption
As a result of this knock out provision, single name default swaptions are rather
illiquid and, typically, only short dated contracts trade.
The knock out feature inserts digital risk into a single name swaption, which may
have unpleasant consequences.
For example, a holder of a payer swaption sees its value go through the roof as
the spread blows out (and the swaption is deep in the money), and drop to zero if
the underlying name defaults prior to T0 .

Default swaptions
Index options
Valuation of default swaptions

Consider a payer swaption (with the knock out provision), i.e. an option to buy
protection, struck at K . Its payoff (exercise value) is
Vpay (T0 , T ) = 1τ >T0 (Vprot (T0 , T ) − K A(T0 , T ))+

= 1τ >T0 A(T0 , T )(C(T0 , T ) − K )+ ,
where C(T0 , T ) = Vprot (T0 , T )/A(T0 , T ) is the forward par spread at T0 .

This expression looks very much like the payoff of an interest rate swaption,
except that for the presence of the indicator function 1τ >T0 (reflecting the knock
out provision).
We can include the knock out feature in the risky annuity by defining
A∗ (t, T0 , T ) = 1τ >t A(t, T0 , T ),
so that the payoff takes the form
Vpay (T0 , T ) = A∗ (T0 , T )(S0 − K )+ .

Default swaptions
Index options
This expression looks exactly like the payoff of an interest rate swaption, and we
are inclined to use the methods of arbitrage pricing theory.
However, while A∗ represents the price of a traded asset (namely a stream of
risky cash payments), strictly speaking, it is not a a numeraire, as it is not
positive.
From a practical perspective, this is not a problem because the value of the
swaption is zero when the numeraire turns zero.
In order to formulate this mathematically, we start with the valuation formula
under the risk neutral measure.

Default swaptions
Index options
Using the results presented in Lecture Notes #2, we find that
h R T0 i
Vpay (0) = E 1τ >T0 A(T0 , T )(C(T0 , T ) − K )+ e− 0 r (s)ds
h R T0 i
= E A(T0 , T )(C(T0 , T ) − K )+ e− 0 (r (s)+λ(s))ds .
In the risk neutral measure, we have thus transformed the (non-strictly

Rt
positive)
indicator process 1t>τ into the (strictly positive) process e− 0 (r (s)+λ(s))ds .

Default swaptions
Index options
R T0
The factor e− 0 (r (s)+λ(s))ds can be combined with A(T0 , T ), yielding the
following numeraire:
N (t) = A(t, T0 , T )
X
= δi P(t, t, Ti ).
1≤i≤N
Note that this numeraire does not correspond to an actual financial asset (as the
risky zero coupon bonds are not modeled correctly), but serves as a useful
“pseudo-asset”.
We let QT0 ,T denote the EMM associated with this numeraire, called the forward
survival measure.

Default swaptions
Index options
Using Girsanov’s theorem and changing to QT0 ,T , we can write the swaption
price in the familiar form
QT ,T
Vpay = A0 (T0 , T )E 0 [(C(T 0, T ) − K )+ ].
Note that, under the survival measure QT0 ,T , the par spread
Vprot (t, T0 , T )
C(t, T0 , T ) =
A(t, T0 , T )
is a martingale.
The precise nature of this martingale is, of course, unknown, and we have to use
approximations.

Default swaptions
Index options
Lognormal model for default swaptions
The market convention for option pricing is to assume that C (t) follows a
lognormal process:
dC (t)
= σdW (t) .
C (t)
As consequence, the swaption price is given by a Black-Scholes type formulas:

Vpay = A0 (T ) C0 N(d1 ) − KN(d2 ) ,

Vrec = A0 (T ) KN(−d2 ) − C0 N(−d1 ) .
where
C0 1
log K
+ 2
σ 2 T0
d1 = p ,
σ T0
C0 1
log K
− 2
σ 2 T0
d2 = p .
σ T0

Default swaptions
Index options
Mechanics of an index option
There is a liquid market for short dated European options on standardized credit
indices.
Since the indices roll every 6 months, and the off-the-run indices are less liquid,
most of the option activity is in the 1 through 3 months.
At expiration, the option holder has the right to exercise into a basket of CDSs at
a specified spread.
The content of the basket consists of the components of the underlying index, at
the time of option trade.

Default swaptions
Index options
Mechanics of an index option
Options on indices do not have the knock out feature.

If there are substitutions or defaults in the basket during the life of the option, this
does not affect the deliverable basket in the option contract.
In particular, all defaulted CDSs should be delivered at option expiration, with
values of one minus recovery.
This is in contrast with single name CDS swaptions.

Default swaptions
Index options
Valuation of an index option
Consider a call option on an index struck at K . For simplicity, we assume that

each component CDS has a notional of $1 (rather than 1/N of the face value).
The payoff of this option is given by
N
X +
j
Vcall (T0 ) = Vprot (T0 , T ) − K
j=1
X +
j
X
= Vprot (T0 , T ) + (1 − Rj ) − K ,
j6∈D j∈D
where D denotes the set of all index components that defaulted prior to T0 .

Default swaptions
Index options

Under the risk neutral measure,
hX i N
hX i
j
E Vprot (T0 , T ) = E 1τj >T0 Aj (T0 , T )(C j (T0 , T ) − C)
j6∈D j=1
N
X
= P0 (T0 )−1 Aj (0, T0 , T )(C j (T0 , T ) − C),
j=1
where C j is the par spread for name j, and C is the spread on the index.
Also,
hX i N
hX i
E (1 − Rj ) = E 1τj <T0 (1 − Rj )
j∈D j=1
N
X
= Qj (0, T0 )(1 − Rj ).
j=1

Default swaptions
Index options
Let us now write the payoff as
Vcall (T0 ) = (Φ(T0 ) − K )+ .
As a result of the calculations above, we know that
N
X
E[Φ(T0 )] = P0 (T0 )−1 Aj (0, T0 , T )(C j (T0 , T ) − C) + Qj (0, T0 )(1 − Rj ).
j=1
To compute the option value Vcall (T0 ), we need more information about the
distribution of Φ(T0 ) beyond its expected value.
This can be done by means of the following approximations.
We assume that Φ(T0 ) can be expressed terms of a suitable “effective” spread
C(T0 ).

Default swaptions
Index options
A simple approach consists in writing:
Φ(T0 ) = p(C(T0 ))
= NAC (T0 , T )(C(T0 ) − C),
where AC (T0 , T ) is the “effective” risky annuity given in terms of the “effective”
survival probabilities SC (T0 , Tj ) introduced below.
The survival probabilities SC (T0 , Tj ) are defined as follows.
We use the credit triangle C0 ≈ λ(1 − R) to write
SC (T0 , Tj ) = e−(C(T0 )/(1−Reff ))(Tj −T0 ) ,
where Reff is an effective recovery rate (usually set to 40%).

Default swaptions
Index options
The expression for SC is similar to the way a CDS index is quoted by assuming
that all names have identical characteristics.
We write the payoff of the option as
+
Vcall (T0 ) = p(C(T0 )) − K ,
and so +
Vcall (0) = P0 (T0 )E p(C(T0 )) − K .

Default swaptions
Index options
If we know the density ϕ of the effective coupon C(T0 ), we can compute this
option value by numerical evaluation of the integral:
Z ∞
Vcall = P0 (T0 ) (p(c) − K )+ ϕ(c)dc.
0
This is subject to the constraint on the expected value:

Z ∞
E p(C(T0 )) = p(c)ϕ(c)dc
0
N
j
X
= P0 (T )−1 Aj (C0 − C) + Qj (0, Tj )(1 − Rj ).
j=1

Default swaptions
Index options
The market convention is to assume the lognormal distribution, and so we write

√ 2
T0 x− 12 σC T0
C(T0 ) = µeσC , where x ∼ N(0, 1).
The index spread volatility σC is a market observable.

The mean µ = E[C(T0 )] is found by enforcing the condition for E p(C(T0 ))
above.

Default swaptions
Index options
To summarize the above calculations, the call price is given by
∞ √
P0 (T )
Z +
2
T x− 12 σC 1 x2

Vcall = √ p µeσC T
−K e− 2 dx,
2π −∞
subject to the condition for µ:
∞ √
1
Z
2 1 2
T x− 12 σC
p µeσC T
e− 2 x dx,

E p(C(T0 )) = √
2π −∞
where E[Φ(T0 )] is defined above.

The integrals can be calculated numerically.
Note that the strike K is often given in spread terms rather than in price terms.
The conversion from spread to strike follows the algorithm presented in Lecture
Notes #3 (which is also similar to the definition of C(T0 ) above).

Default swaptions
Index options
CMDS
A CMDS is a variation on the CDS in which one leg pays a periodic floating
(rather than fixed) coupon.
This coupon is linked to the fixing of a reference spread on the previous coupon
date.
Denote the coupon dates on the swap by Ti , i = 1, . . . , M. On the date Ti the
coupon leg pays the par yield C(Ti−1 ) of a reference CDS maturing, say, 5 years
later.
Since the maturity of the reference CDS remains constant throughout the life of
the contract, it i referred to as a constant maturity default swap.

Default swaptions
Index options
CMDS
Typically, the other leg on the swap is the standard protection leg.
It may also be a standard premium leg, or a CMDS leg corresponding to a
different maturity CDS (say, 10 years).
To develop a pricing model, consider first a single coupon setting at time T0 , on a
par rate corresponding to a CDS with payment dates T1 , . . . , TN = T , so that the
maturity of the reference swap is T − T0 . The CMDS coupon is paid on T1 .
For simplicity, we assume that there is no payment at time T1 if τ < T1 (no
accrued interest).

Default swaptions
Index options
Valuation of a CMDS
The value of this payment on the settlement date is
Vcpn (T0 ) = 1τ >T0 P(T0 , T1 )C(T0 )δ,
where δ is a day count fraction.

Today’s value of the payment is
R T0
Vcpn (0) = E[1τ >T0 P(T0 , T1 )C(T0 )e− 0
r (s)ds
]δ
Q −1
= A(0)E T0 ,T [P(T 0 , T1 )A(T0 ) C(T0 )]δ,
where QT0 ,T is the survival measure.

We know that C(t) is a martingale under the measure QT0 ,T .
P(t,T1 )
The quantity A(t)
is also a martingale.

Default swaptions
Index options
Andersen’s method
Recall the property of iterated expected value:
E[E[X |Y ]] = E[X ].
Following Leif Andersen, we use this identity to write
QT ,T
Vcpn (0) = A(0)E 0 [f (C(T 0 ))C(T0 )]δ,
where f (C(T0 )) is the conditional expectation:
QT ,T −1
f (C(T0 )) , E 0 [P(T 0 , T1 )A(T0 ) | C(T0 )].

Default swaptions
Index options
Andersen’s method
Our goal is to find an effective approximation to f (x). Then we set
g (x) = f (x) x,
and compute Vcpn (0) by integration:

Z ∞
Vcpn (0) = A(0)δ g (x) ϕ (x) dx,
0
where ϕ is the probability density of C(T0 ).

The density ϕ is often known from the assumed model.
It can also be determined from the observed swaption prices by means of the
replication formula, which we discussed in Lecture Notes #6.

Default swaptions
Index options
Andersen’s method
Therefore
∂2
ϕ(K ) = A(0)−1 Vcall (K ),
∂K 2
so that Z ∞ ∂2
Vcpn (0) = δ g (K ) Vcall (K ) dK .
0 ∂K 2
This shows that the CMDS payout can be replicated by taking positions in CDS
swaptions at many different strikes.
The weight on the swaption with strike K is g 00 (K ).

Default swaptions
Index options
Andersen’s method
We still need to find an approximate form of f .

To this end, we notice that the graph of P(T0 , T1 )A(T0 )−1 against C(T0 ) is
roughly a straight line (in reality, the ratio depends on the entire term structure of
spreads...).
We thus write
f (C) ≈ aC + b.
To determine the constant b, we can use the fact that for C = 0,
P(T0 , T1 )A(T0 )−1 = P(T0 , T1 )A(T0 )−1 ,
which implies that

b = P0 (T1 )A−1
0 .

Default swaptions
Index options
Andersen’s method
We can then find a from the basic requirement that P(t, T1 )/A(t) be a
martingale in the annuity measure, such that
QT ,T
E 0 [aC(T 0) + b] = aC0 + b
= P0 (T1 )/A(0).
or
P0 (T1 )/A(0) − b
a= .
C0
So, as a reasonable approximation, we have the formula
Z ∞
Vcpn (0) = A(0)δ (ax 2 + bx)ϕ(x)dx.
0
This integral can be carried out analytically in various models.

It can also, in principle, be computed model free, by means of the replication
formula.

Default swaptions
Index options
Pricing CMDS with a cap
The contract we have considered so far is somewhat simplified. For distressed

names, their par spreads can reach very high levels and the CDSs will stop
trading.
As a consequence, it is a standard practice to impose a cap on the forward
spread. Also, there is typically a notional scale α on the coupon leg, to allow the
total CMDS contract to be issued at par, i.e. at zero value.
The coupon payment at time Ti is then
αδ1τ >Ti min(C(Ti−1 ), U),
where U is the cap level.

This change can be easily accommodated: we replace the unconstrained payoff
with
g (x) = αf (x) min(x, U),
where f (x) is the same function as before.

Default swaptions
Index options
Pricing CMDS with a cap
Using Andersen’s method, this yields

Z ∞
Vcpn (0) = αδA(0) (ax + b) min(x, U)ϕ (x) dx
0
Z U Z ∞
= αδA(0) (ax 2 + bx)ϕ (x) dx + αδA(0)U (ax + b)ϕ (x) dx.
0 U
For CMDS contracts, the value of α that renders the values of both legs of the
swap identical is denoted αpar ; this is the quoted value.
For the lognormal model, CMDS caps can be priced in closed form.

Default swaptions
Index options
References
Andersen, L.: Credit models, lecture notes (2010).
O’Kane, D.: Modelling Single-Name and Mult-Name Credit Derivatives, Wiley

(2008).
Schönbucher, P. J.: Credit Derivatives Pricing Models, Wiley Finance (2003).

7 - Default Swaptions, Index Options, and CMDS

Uploaded by

Copyright:

Available Formats

7 - Default Swaptions, Index Options, and CMDS

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

7 - Default Swaptions, Index Options, and CMDS

Uploaded by

Copyright:

Available Formats

Default swaptions

Interest and Credit Models

A. Lesniewski Interest and Credit Models

3 Constant maturity default swaps (CMDS)

A. Lesniewski Interest and Credit Models

A. Lesniewski Interest and Credit Models

A. Lesniewski Interest and Credit Models

Valuation of default swaptions

Vpay (T0 , T ) = 1τ >T0 (Vprot (T0 , T ) − K A(T0 , T ))+

where C(T0 , T ) = Vprot (T0 , T )/A(T0 , T ) is the forward par spread at T0 .

A∗ (t, T0 , T ) = 1τ >t A(t, T0 , T ),

so that the payoff takes the form

Vpay (T0 , T ) = A∗ (T0 , T )(S0 − K )+ .

A. Lesniewski Interest and Credit Models

Valuation of default swaptions

A. Lesniewski Interest and Credit Models

Valuation of default swaptions

Using the results presented in Lecture Notes #2, we find that

In the risk neutral measure, we have thus transformed the (non-strictly

A. Lesniewski Interest and Credit Models

Valuation of default swaptions

A. Lesniewski Interest and Credit Models

Valuation of default swaptions

A. Lesniewski Interest and Credit Models

Lognormal model for default swaptions

A. Lesniewski Interest and Credit Models

Mechanics of an index option

A. Lesniewski Interest and Credit Models

Mechanics of an index option

Options on indices do not have the knock out feature.

A. Lesniewski Interest and Credit Models

Valuation of an index option

Consider a call option on an index struck at K . For simplicity, we assume that

A. Lesniewski Interest and Credit Models

Valuation of an index option

A. Lesniewski Interest and Credit Models

Valuation of an index option

Let us now write the payoff as

Vcall (T0 ) = (Φ(T0 ) − K )+ .

As a result of the calculations above, we know that

A. Lesniewski Interest and Credit Models

Valuation of an index option

A simple approach consists in writing:

SC (T0 , Tj ) = e−(C(T0 )/(1−Reff ))(Tj −T0 ) ,

where Reff is an effective recovery rate (usually set to 40%).

A. Lesniewski Interest and Credit Models

Valuation of an index option

A. Lesniewski Interest and Credit Models

Valuation of an index option

This is subject to the constraint on the expected value:

A. Lesniewski Interest and Credit Models

Valuation of an index option

The market convention is to assume the lognormal distribution, and so we write

The index spread volatility σC is a market observable.

A. Lesniewski Interest and Credit Models

Valuation of an index option

To summarize the above calculations, the call price is given by

subject to the condition for µ:

where E[Φ(T0 )] is defined above.

A. Lesniewski Interest and Credit Models