Ratchet Cap-Floors Bocconi4

Copyright 2005 Damiano Brigo: The LIBOR and Swap market models
Banca IMI
LECTURE 4
LECTURE 4 Pricing with the LIBOR model: Libor in Arrears; Constant Maturity Swaps (CMS); Ratchet caps and oors; Zero Coupon Swaptions.
Universit` Bocconi: Fixed Income a
162
Banca IMI
LIBOR model pricing: General approximation

Freeze part of the drift of the LIBOR dynamics so as to obtain a multi-dimensional geometric Brownian motion. This was done earlier to derive approximated formulas for swap volatilities and terminal correlations. Recall: under the Ti-forward-adjusted measure Qi we have the exact dynamics:
dFk (t) = i,k (t)Fk (t) dt + k (t)Fk (t) dZk (t) ,

where i,k (t) := k (t)k (t) for i < k, i,i(t) := 0 and i i i,k (t) := k (t)k (t) for i > k (see Lecture 1). To sum up:
i,k (t) i,k (t) i,k (t)
:= := :=
k (t)
i X k,j j j (t)Fj (t) j=k+1
1 + j Fj (t)
, k<i
0, k = i k (t)
k X k,j j j (t)Fj (t) j=i+1
1 + j Fj (t)
, k > i.
The distributions or statistical laws of the Fk under Qi are unknown for i = k. This is a problem, because prices are expectations under pricing measures, and if we do not know the laws of the random variables we cannot compute the expectations analytically. We are forced to resort to numerical methods. Can we escape this situation in some cases?
Universit` Bocconi: Fixed Income a 163
Banca IMI

i X k,j j j (t)Fj (t) j=k+1
i,k (t) i,k (t) i,k (t)
:= := :=
k (t)
1 + j Fj (t)
, k<i
0, k = i k (t)
k X k,j j j (t)Fj (t) j=i+1
1 + j Fj (t)
, k > i.
Consider the approximated lognormal dynamics:

i X k,j j j (t)Fj (0) j=k+1
i,k (t) i,k (t) i,k (t)
:= := :=
k (t)
1 + j Fj (0)
, k<i
0, k = i k (t)
k X k,j j j (t)Fj (0) j=i+1
1 + j Fj (0)
, k > i.
164
Banca IMI
Consider the approximated lognormal dynamics:

i X k,j j j (t)Fj (0) j=k+1
i,k (t) i,k (t) i,k (t)
:= := :=
k (t)
1 + j Fj (0)
, k<i
0, k = i k (t)
k X k,j j j (t)Fj (0) j=i+1
1 + j Fj (0)
, k > i.
This dynamics gives access, in some cases, to a number of techniques which have been developed for the basic Black and Scholes setup, for example, in equity and FX markets. Moreover, this freezing-part-of-the-drift technique can be combined with drift interpolation so as to allow for rates that are not in the fundamental (spanning) family T0, T1, . . . , TM corresponding to the particular model being implemented. Finally, even resorting to MC allows now for a one-shot propagation of the dynamics with no infra-discretization, thus reducing memory requirements and simulation time. A similar idea may work also in some smile extensions.
Banca IMI
Libor in Arrears (In Advance Swaps)
An in-advance swap (or LIBOR in arrears) is an IRS that resets at dates T+1, . . . , T and pays at the same dates, with unit notional amount and with xed-leg rate K . With respect to standard swaps, the LIBOR payments are in arrears, since the libor pays immediately when it resets, and not one period later. More precisely, the discounted payo of an in-advance swap (of payer type) can be expressed via
X B(0)
B(Ti) i=+1 =
i+1(L(Ti, Ti+1) K) =
X B(0)
B(Ti) i=+1
i+1(Fi+1(Ti) K).
The value of such a contract is, therefore,
2
IAS = E 4
B
B(0) i+1(Fi+1(Ti) K)5 . B(Ti) i=+1
166
Banca IMI
Libor in Arrears (In Advance Swaps)
Before calculating the expectations, it is convenient to make some adjustments. We shall use the following identity (obtained easily via iterated conditioning):
E =E =E

B(0) XT B(T )

=E
P (T, S) B(0) XT P (T, S) B(T )

= =
1 B(0) B B(T ) XT E 1 InfoT P (T, S) B(T ) B(S) E

B

=
1 B(0) B(T ) XT 1 InfoT P (T, S) B(T ) B(S)
=E
1 B(0) InfoT XT P (T, S) B(S)

=
so that
=E
1 B(0) XT P (T, S) B(S)

XT
B(0) B(T )
2
B =E 4
B(0) XT B(S)
3 5
for all 0 < T < S,
P (T, S)
where X is a T -measurable random variable, known from InfoT

Banca IMI
To value the above contract, notice that
8 9 < X B(0) = E (F (T ) K) :i=+1 B(Ti) i+1 i+1 i ; 8 9 = <X 1 = ... (1 + i+1K) =E D(0, Ti) ; :i=+1 P (Ti, Ti+1)
Now use our previous result with T = Ti, S = Ti+1, XT = 1/P (Ti, Ti+1) to get

E
1 B(0) B(Ti) P (Ti, Ti+1)
2
=E4
B(0) 1 P (Ti ,Ti+1 ) B(Ti+1 )
3 5=
P (Ti, Ti+1)
1 B(0) =E P (Ti, Ti+1)2 B(Ti+1)

and substitute:
8 9 = <X B(0) 1 B(0) =E (1 + i+1K) :i=+1 B(Ti+1) P (Ti, Ti+1)2 B(Ti) ;
168
Banca IMI
=E
X B(0) i=+1
1 B(0) (1 + i+1K) B(Ti+1) P (Ti, Ti+1)2 B(Ti)
2
B 6 6 B(0) B(Ti+1)
3 7 1 7 25 P (Ti, Ti+1) 3
1
P (Ti+1 , Ti+1 )
X i=+1
X i=+1
B(0) (1 + i+1K) B(Ti)

i+1
2
P (0, Ti+1 )
X i=+1
6 6 4
7 1 7 25 P (Ti, Ti+1)
X i=+1
P (0, Ti)(1 + i+1K)
X i=+1
P (0, Ti+1)E
i+1
(1 + i+1Fi+1(Ti))
X i=+1
P (0, Ti)(1 + i+1K).
169
Banca IMI
Computing the expected value is an easy task, since we know that, under Qi+1, Fi+1 has the driftless (martingale) lognormal dynamics
dFi+1(t) = i+1(t)Fi+1(t)dZi+1(t) ,
so that (Ito formula (F ) = F 2, (F ) = 2F, (F ) = 2),
dFi+1(t) = 2Fi+1(t)dFi+1(t) + 1 2dFi+1(t)dFi+1(t) 2 = i+1(t) Fi+1(t)dt + 2i+1(t)Fi+1(t)dZi+1(t) ,

so that we still have a geometric brownian motion for F 2:
2 2 2
dFi+1(t) = i+1(t) Fi+1(t)dt + 2i+1(t)Fi+1(t)dZi+1(t) ,

and the mean of this process is known to be
i+1
Fi+1(Ti)
= Fi+1(0) exp
2
"Z
0
Ti
#
i+1(t)dt
2 2
= Fi+1(0) exp(Tivi )
where the v s have been dened in Lecture 2 and are caplet volatilities for Ti Ti+1.
170
Banca IMI
By expanding the square and substituting we obtain IAS
= +
X i=+1
{P (0, Ti+1) [1 + 2i+1Fi+1(0)+
2 2 2 i+1Fi+1(0) exp(vi Ti)
(1 + i+1K)P (0, Ti)}.
Contrary to the plain-vanilla case, this price depends on the volatility of forward rates through the caplet volatilities v . Notice however that correlations between dierent rates are not involved in this product, as one expects from the additive and one-rateper-time nature of the payo.
171
Banca IMI
Ratchet Caps and Floors

A ratchet cap is a cap where the strike is updated at each caplet reset, based on the previous realization of the relevant interest rate. A simple ratchet cap rst resetting at T and paying at T+1, . . . , T pays the following discounted payo:
X i=+1
D(0, Ti)i [L(Ti1, Ti) (L(Ti2, Ti1) + X)] ,
Notice that if we set Ki := L(Ti2, Ti1) + X for all is this is a set of caplets with (random) strikes Ki.
X is a margin, which can be either positive or negative.

A sticky ratchet cap is instead given by:
X i=+1
D(0, Ti)i [L(Ti1, Ti) Xi] ,
Xi = max L(Ti2, Ti1) X, Xi1 X , X := L(T1, T).

There are versions with min replacing max in the Xis denition. The quantity X is a spread that can be positive or negative.
Banca IMI
Ratchet caps and oors
In general a sticky ratchet cap has to be valued through Monte Carlo simulation. We have
E{
D(0, Ti)i [L(Ti1, Ti) Xi] }
i=+1 X i=+1
(
i E
= P (0, T )
[L(Ti1, Ti) Xi] P (Ti, T )
)
.
Since the Q forward-rate dynamics of F(t)(t), . . . , F (t) can be discretized via the usual scheme, Monte Carlo pricing can be carried out in the usual manner. We can use also the lognormal frozen-drift approximation to implement a faster MC simulation. However, for the non-sticky ratchet cap payo we may investigate possible analytical approximations based on the usual freezing the drift technique for the LIBOR market model.
173
Banca IMI
Non-Sticky Ratchets: Analytical approximation
All we need to compute is the expectation
E{D(0, Ti) [L(Ti1, Ti) (L(Ti2, Ti1) + X)] } = P (0, Ti)E {[Fi(Ti1) Fi1(Ti2) X] } =: P (0, Ti)mi and then add terms. In the above expectation, the rates evolve as follows under the measure Qi: dFi(t) = i(t)Fi(t)dZi(t), dFi1(t) = i1,iii(t)Fi(t) Fi1(t)dt + i1(t)Fi1(t)dZi1(t) 1 + iFi(t)
i +
As usual, in such dynamics we do not know the distribution of Fi1(t). But, since Fi1 and Fis reset times are adjacent, we may freeze the drift in Fi1 and be rather condent on the resulting approximations. We thus replace the second SDE by
dFi1(t) (t)
= :=
(t)Fi1(t)dt + i1(t)Fi1(t)dZi1(t), i1,iii(t)Fi(0) . 1 + iFi(0)
Now both Fi1 and Fi follow (correlated) geometric Brownian motions as in the Black and Scholes model.
174
Banca IMI
Non-Sticky Ratchets: Analytical approximation. Spread option

Now consider the case X > 0. If we set S1 := Fi, S2 := Fi1, r q1 := 0, r q2 := (t)1{t < Ti2}, 1 := i(t), 2 := i1(t)1{t < Ti2}, a = 1, b = 1, and = 1, we may view our dynamics as the two-dimensional Black Scholes dynamics d[S1, S2] and our payo as a spread option payo, by slightly adjusting to the fact that no discounting should occur in our case. Consider two assets whose prices S1 and S2 evolve, under the risk neutral measure, according to
dS1(t) = S1(t)[(r q1)dt + 1dW1 (t)], S1(0) = s1, dS2(t) = S2(t)[(r q2)dt + 2dW2 (t)], S2(0) = s2,
Q Q where W1 and W2 are Brownian motions under Q with instantaneous correlation . Fix a maturity T , a positive real number a, a negative real number b, a strike price K . The spread-option payo at time T is then dened by Q
H = (awS1(T ) + bwS2(T ) wK) ,

where w = 1 for a call and w = 1 for a put.
(9)
175
Banca IMI
Non-Sticky Ratchets: Analytical approximation. Spread option

Price of the spread option:
r(T t) Q Et
t = e
(awS1(T ) + bwS2(T ) wK)
A pseudo-analytical formula can be derived. The unique arbitragefree price is
Z
t =
1 1 v2 e 2 f (v)dv, 2
where
1 2 2 f (v) = awS1(t) exp q1 1 + 1 v 2 0 1 aS1 (t) 2 2 1 ln h(v) + [1 + ( 2 )1 ] + 1 v A @w p 1 1 2
0
wh(v)e
and
r
@w
ln
aS1 (t) h(v)
+ (1 p 1 1 2
1 2 2 1 )
1 + 1 v
h(v) = KbS2(t)e
2 (2 1 2 ) +2 v 2
, 1,2 = rq1,2, = T t.
proof based on standard bivariate Gaussian variables comp.

Banca IMI
Non-Sticky Ratchets: Analytical approximation.

If X < 0 we just switch the denitions of S1 and S2 above, S1 := Fi1, S2 = Fi etc., and then take = 1. In the calculations below we assume X > 0. As a matter of fact, our coecients here are time-dependent, but this does not change substantially the derivation. It follows that our expected value
mi := E {[Fi(Ti1) Fi1(Ti2) X] }
is given by our formula above for the spread option when taking into account the above substitutions, i.e. one needs to apply said formula with a = 1, = 1, t = 0,
Z
S1(t) = Fi(0), q1 = 0, q2 = r = 0,
2 1
Ti2 0
(u)du,
Ti2
Z
=
0
Ti1
2 i (u)du,
2 2
Z
0
i1(u)du,
= i1,i, K = X.
Once we have the mis, our ratchet price is given by
X i=+1
P (0, Ti)imi.
177
Banca IMI
Non-Sticky Ratchets: Analytical approximation. Case X = 0.

The above price depends on a one-dimensional numerical integration. There is a case, though, where this is not necessary. Indeed, if X = 0, we obtain a special ratchet cap that, under the lognormal assumption, we may value analytically through the Margrabe formula for the option exchanging one asset for another. We map the ratchet payo terms
(Fi(Ti1) Fi1(Ti2))
into equity payos
H = (S1(T ) S2(T ))
This payo is the so called option to exchange one asset (S1) for another (S2). Indeed, if we hold S2, when we are at T the option pays
(S1(T )S2(T )) = max(S1(T )S2(T ), 0) = S1(T )S2(T )

if S1(T ) > S2(T ), and 0 otherwise. Recall we are holding S2. By getting the option payo in this case where S1(T ) > S2(T ), we get a total of
S2(T ) + (S1(T ) S2(T )) = S1(T ).

Banca IMI
So, when holding S2, if S1(T ) > S2(T ) by means of option we end up with S1, so we have exchanged S2 with more valuable S1. On the contrary, if S1(T ) < S2(T ), option expires worthless and there is no exchange, so we keep more valuable S2.
the the the the
This means that the exchange is a right but no obligation, since it occurs only when it favors us. This is then indeed an option to exchange one asset for another. In the market this kind of option is priced with Margrabes formula, which we derive below by using the change of numeraire technique.
179
Banca IMI
We derive a formula now for
B(0) S1(0) + S + (S1(T ) S2(T )) =E 1 (S1(T ) S2(T )) = B(T ) S1(T )

S (0)E 1
"
= S1
where Yt = Note that we took S1 as numeraire, assuming q1 = 0, since the numeraire has to be a positive non-dividend paying asset (q1 = 0). Notice also that in the numerator, to have the price of a tradable asset, we got rid of the dividend by inserting the forward price
RT t q2 (s)ds S2(t)e /S1(t).
S2(T ) 1 S1(T )
+ #
= S1(0)E
S1
(1 Y (T ))
Et
RT B(t) t q2 (s)ds S2(T ) = S2(t)e B(T )
(without dividends the forward price would be S2(t) itself). Now we need to derive the dynamics of Yt under the S1 measure. We know this is a martingale, since Yt is a ratio between a tradable asset and our numeraire S1, so that by FACT ONE on the change of numeraraire (rst lecture) we have that Yt is a martingale (=zero drift).
Banca IMI
Compute then, rst under QB :
S2(t)e t q2(s)ds A=d dYt = d @ S1(t)
RT
RT
q2 (s)ds S2 (t) S1(t)

=
First notice that the rst term would only give a dt contribution when dierentiated. Then we compute directly
=
RT
RT t q2 (s)ds e d
S2(t) S1(t)

+ (. . .)dt =
1 1 q (s)ds =e t 2 d (S2(t)) + S2(t)d + S1(t) S1(t) 1 + (. . .)dt = + dS2(t) d S1(t) RT 1 1 q (s)ds =e t 2 d (S2(t)) + S2(t)d +(. . .)dt = S1(t) S1(t) RT 1 q (s)ds B =e t 2 S2(t)[(r q2)dt + 2dW2 ]+ S1(t) 1 + S2(t)d + (. . .)dt = ... S1(t) 1 1 Since (Ito (S) = S , (S) = S 2 , (S) = 2/(S 3))

d
1 S1(t)

=
1 1 2 B 1 dS1+ 2 3 dS1dS1 = [rdt+1dW1 ]+(. . .)dt, 2 S1 S1 S1

181
Banca IMI
by substituting we obtain
... = e
t q2 (s)ds
RT
1 B S2(t)2dW2 + S2(t) S1(t) +(. . .)dt =

R
1 B 1dW1 S1

Recalling that Yt = S2e
q2
/S1, we may then write

B
dYt = Yt1dW1 + Yt2dW2 + (...)dt

Now, if we change numeraire, the diusion part does not change. Since we already know that under the S1 measure Y is a martingale, this means that the equation of Y under S1 will have the same diusion parts and zero drift. We get
dYt = Yt1dW1 1 + Yt2dW2 1

or also
dYt = Yt(1dW1 1 + 2dW2 1 )

From the point of view of the law, this process is the same as a process with a single brownian motion
dYt = Yt(0dW0 1 )
provided that Var(2dW2 1 1dW1 1 ) = Var(0dW0 1 )
Banca IMI
This equation reads Var(2dW2 1 1dW1 1 ) = 1 dt + 2 dt 212dt and since we have Var(0dW0 1 ) = 0 dt
S 2 S S 2 2
0 = 1 + 2 212
Let us now go back to
S1(0)E
with the dynamics
S1
(1 Y (T ))
dYt = Yt(0dW0 1 )
This is a put option with strike 1 for which we get the formula
S1(0)[(d2) Y (0)(d1)]
with
d1,2 =
ln(Y (0)/1)
R T
0
1 2
RT
0
2 0 (t)dt
1
2
2 0 (t)dt
Recalling the expressions for Y and 0 we get
[S1(0)(d2) S2
RT 0 q2 (t)dt (0)e (d
1 )]
183
d1,2 =
ln(S2(0)/S1(0))
RT
0
q2(t)dt
R T
0
1 2
RT
0
Banca IMI
[...]dt
2 2 [1 (t) + 2 (t) 21(t)2(t)]dt
1
2
As before, set
Z
S1(t) = Fi(0), q1 = 0,
Z
q2(t)dt =
0
Ti2
(u)du,
Z
r = 0,
T 0
2 1 (t)dt
Z
=
0 2
0 Ti1
i (t)dt,
Z
0
2 2 (t)dt
Z
=
0
Ti2
i1(t)dt, = i1,i
to get:
184
Banca IMI
8 < X D(0, Ti )i L(Ti1 , Ti ) L(Ti2 , Ti1 ) E : 8 < X =E D(0, Ti )i Fi (Ti1 ) Fi1 (Ti2 ) :
i=+1 i=+1 +
9 = ;
9 = ; ! #
(d2 ) ,
i
X i=+1
"
i P (0, Ti ) Fi (0)(d1 ) Fi1 (0) exp
i
Z
0
Ti2
(u)du
RT ln(Fi (0)/Fi1 (0)) 0 i2 (u)du i 1 d1,2 = 2 Ri , Ri Z

Ri =
0 Ti1 2 i (u)du +
Z
0
Ti2
!1
2 (i1 (u) 2i1,i i1 (u)i (u))du
In this section we dealt with ratchet caps. The treatment of ratchet oors is analogous.
185
Banca IMI
Zero Coupon Swaption
A payer (receiver) zero-coupon swaption is a contract giving the right to enter a payer (receiver) zero-coupon IRS at a future time. A zero-coupon IRS is an IRS where a single xed payment is due at the unique (nal) payment date T for the xed leg in exchange for a stream of usual oating payments iL(Ti1, Ti) at times Ti in T+1, T+2, . . . , T (usual oating leg). In formulas, the discounted payo of a payer zero-coupon IRS is, at time t T:
B(t) 4 P (T, Ti)iFi(T) P (T, T ), K 5 , B(T) i=+1

where , is the year fraction between T and T . The analogous payo for a receiver zero-coupon IRS is obviously given by the opposite quantity. Taking risk-neutral expectation, we obtain easily the contract value as
P (t, T) P (t, T ) , KP (t, T ),

which is the typical value of a oating leg minus the value of a xed leg with a single nal payment.
Banca IMI
The value of K that renders the contract fair is obtained by equating to zero the above value. K = F (t; T, T ). Indeed, the value of the swap is independent of the number of payments on the oating leg, since the oating leg always values at par, no matter the number of payments. Therefore, we might as well have taken a oating leg paying only in T the amount , L(T, T ). This would have given us again a standard swaption, standard in the sense that the two legs of the underlying IRS have the same payment dates (collapsing to T ) and the unique reset date T. In such a one-payment case, the swap rate collapses to a forward rate, so that we should not be surprised to nd out that the forward swap rate in this particular case is simply a forward rate.
187
Banca IMI
An option to enter a payer zero-coupon IRS is a payer zerocoupon swaption, and the related payo is
B(t) X + [ P (T, Ti)iFi(T) P (T, T ), K] , B(T) i=+1
or, equivalently, by expressing the F s in terms of discount factors,
B(t) + [1 P (T, T ) P (T, T ), K] , B(T)

which in turn can be written as
B(t) + , P (T, T ) [F (T; T, T ) K] . B(T)
188
Banca IMI
B(t) + , P (T, T ) [F (T; T, T ) K] . B(T)

Notice that, from the point of view of the payo structure, this is merely a caplet. As such, it can be priced easily through Blacks formula for caplets. The problem, however, is that such a formula requires the integrated percentage variance (volatility) of the forward rate F (; T, T ), which is a forward rate over a nonstandard period. Indeed, F (; T, T ) is not in our usual family of spanning forward rates, unless we are in the trivial case = +1. Therefore, since the market provides us (through standard caps and swaptions) with volatility data for standard forward rates, we need a formula for deriving the integrated percentage volatility of the forward rate F (; T, T ) from volatility data of the standard forward rates F+1, . . . , F . The reasoning is once again based on the freezing the drift procedure, leading to an approximately lognormal dynamics for our standard forward rates.
189
Banca IMI

Denote for simplicity F (t) := F (t; T, T ) and := , . We begin by noticing that, through straightforward algebra, we have (write everything in terms of discount factors to check)
1 + F (t) =
It follows that
Y j=+1
(1 + j Fj (t)).
ln(1 + F (t)) =
so that d ln(1 + F (t)) =
X j=+1
ln(1 + j Fj (t)),
X j=+1
d ln(1 + j Fj (t)) =
j dFj (t) + (. . .)dt. 1 + j Fj (t) j=+1
1 + F (t) d ln(1 + F (t)) + (. . .)dt, we obtain from the above expression

Since dF (t) =
1 + F (t) X j dFj (t) dF (t) = + (. . .)dt. 1 + j Fj (t) j=+1
Banca IMI
1 + F (t) X j dFj (t) dF (t) = + (. . .)dt. 1 + j Fj (t) j=+1
Take variance (conditional on t) on both sides:

Var
dF (t) F (t)

=
1 + F (t) F (t)
2 X ij i,j i(t)j (t)Fi(t)Fj (t)

i,j=+1
(1 + iFi(t))(1 + j Fj (t))
dt.
Freeze all t s to 0 except for the s, and integrate over [0, T]: zc (v, )2 := (1/T)
1 + F (0) F (0)
2 X
ij i,j Fi(0)Fj (0) (1 + iFi(0))(1 + j Fj (0)) i,j=+1
Z
0
i(t)j (t)dt.
To price the zero-coupon swaption it is then enough to put this quantity into the related Blacks Caplet formula: ZCPS = P (0, T )[F (0)(d1(F (0), K, v, ))
zc
K(d2(F (0), K, v, ))].

zc
Banca IMI
We have checked the accuracy of this formula against the usual Monte Carlo pricing based on the exact dynamics of the forward rates. In the tests all swaptions are at-the-money. We have done this under a number of situations , corresponding to possible modications of the data coming from a standard calibrations of the LIBOR model to at-the-money swaptions data. All cases show the formula to be suciently accurate for practical purposes. When using the formula we notice that the at-the-money standard swaption has always a lower volatility (and hence price) than the corresponding at-the-money zero-coupon swaption. We may wonder whether this is a general feature. Indeed, we have the following.
192
Banca IMI
Comparison between zero-coupon swaptions and corresponding standard swaptions: A rst remark is due for a comparison zc between the zero-coupon swaption volatility v, and the LFM corresponding European-swaption approximation v, . If we rewrite the latter as
LFM 2 T(v, )
X i,j=+1
Z
i,j ij
0
i(t)j (t)dt, i =
wi(0)Fi(0) , S, (0)
it is easy to check that

zc 2 T(v, )
X i,j=+1
Z
i,j ij
0
i(t)j (t)dt,
where
P (0, T) i i , P (0, Ti) the discrepancy increasing with the payment index i. It follows that, for positive correlations, the zero-coupon swaption volatility is always larger than the corresponding plain vanilla swaption volatility, the dierence increasing with the tenor T T, for each given T. i =
193
Banca IMI
Constant Maturity Swaps (CMSs)
A constant-maturity swap is a nancial product structured as follows. We assume a unit nominal amount. Let us denote by {T0, . . . , Tn} a set of payment dates at which coupons are to be paid. At time Ti1 (in some variants at time Ti), i 1, institution A pays to B the c-year swap rate resetting at time Ti1 in exchange for a xed rate K . Formally, at time Ti1 institution A pays to B
Si1,i1+c(Ti1) i ,
instead of
L(Ti1, Ti)i = Fi(Ti1) i , as would be natural (standard Interest Rate Swap with model independent valuation, see Lecture 1).
194
Banca IMI
Constant Maturity Swaps (CMSs)

The net value of the contract to B at time 0 is
E
n X i=1
n X i=1
B(0) (Si1,i1+c(Ti1) K)i B(Ti1)
iP (0, Ti1) E
i1
(Si1,i1+c(Ti1)) K
2
n X i=1
3
B(0) B(Ti1)
i E
B 6 6
7 Si1,i1+c(Ti1)7 5
n X i=1
K
We can change numeraire in two ways:
iP (0, Ti1)
1:
n X i=1
iP (0, Ti1) E
i1
(Si1,i1+c(Ti1)) K
2:
n X i=1

i P (0, Tn) E
Si1,i1+c(Ti1) P (Ti1, Tn)
KP (0, Ti1)) .
195
Banca IMI
CMSs
We need only compute either
i1
[Si1,i1+c(Ti1)] or E [Si1,i1+c(Ti1)/P (Ti1, Tn)]
At rst sight, one might think to discretize the dynamics of the forward swap rate in the swap model under the relevant forward measure, and compute the required expectation through a Monte Carlo simulation. However, notice that forward rates appear in the drift of such equation, so that we are forced to evolve forward rates anyway. As a consequence, we can build forward swap rates as functions of the forward LIBOR rates obtained by the Monte Carlo simulated dynamics of the LIBOR model. Find the swap rate Si1,i1+c(Ti1) from the Ti1 values of the (Monte Carlo generated) spanning forward rates
Fi(Ti1), Fi+1(Ti1), . . . , Fi1+c(Ti1).

Analogously to earlier cases, such forward rates can be generated according to the usual discretized (Milstein) dynamics based on Gaussian shocks and under the unique measure Qn for example.
196
Banca IMI
CMSs
Alternatively, resort to S, (T) and compute
P
i=+1
wi(0)Fi(T)
E S, (T)
X i=+1
X i=+1
wi(0)E Fi(T)
wi(0)e
R T
0
,i (t)dt
Fi(0)
E S, (T)
X i=+1
X i=+1
wi(0)E Fi(T)
0 ,i (t)dt
wi(0)e
R T
Fi(0)
We have frozen again the drift in the Fis dynamics of the F s under Q. This can be compared with classical market convexity adjustments. The two methods give similar results when volatilities are not too high. Notation for was given at the beginning of this unit. The method is general and can be used whenever swap rates or forward rates are paid at times that are not natural in swaps and similar contracts. A dynamics can be obtained by the freezing procedures outlined above.
Banca IMI
Thats all Folks!!!
For questions, further references, etc www.damianobrigo.it damiano.brigo@gmail.com For the exam, it is highly recommended to have a look at the exercises done during the course. For further developments or to expand on the course contents the second edition of Interest Rate Models: Theory and Practice, With Smile, Ination and Credit is a good reference. http://www.damianobrigo.it/book.html
Thanks for your attention and all the best!!
198

Ratchet Cap-Floors Bocconi4

Transféré par

Informations du document

Description originale:

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Ratchet Cap-Floors Bocconi4

Transféré par

Droits d'auteur :

Formats disponibles

Copyright 2005 Damiano Brigo: The LIBOR and Swap market models

Universit` Bocconi: Fixed Income a

LIBOR model pricing: General approximation

dFk (t) = i,k (t)Fk (t) dt + k (t)Fk (t) dZk (t) ,

i,k (t) i,k (t) i,k (t)

i X k,j j j (t)Fj (t) j=k+1

LIBOR model pricing: General approximation

dFk (t) = i,k (t)Fk (t) dt + k (t)Fk (t) dZk (t) ,

i,k (t) i,k (t) i,k (t)

Consider the approximated lognormal dynamics:

dFk (t) = i,k (t)Fk (t) dt + k (t)Fk (t) dZk (t) ,

i,k (t) i,k (t) i,k (t)

Universit` Bocconi: Fixed Income a

LIBOR model pricing: General approximation

Consider the approximated lognormal dynamics:

dFk (t) = i,k (t)Fk (t) dt + k (t)Fk (t) dZk (t) ,

i,k (t) i,k (t) i,k (t)

Libor in Arrears (In Advance Swaps)

The value of such a contract is, therefore,

B(0) i+1(Fi+1(Ti) K)5 . B(Ti) i=+1

Universit` Bocconi: Fixed Income a

Libor in Arrears (In Advance Swaps)

P (T, S) B(0) XT P (T, S) B(T )

1 B(0) B B(T ) XT E 1 InfoT P (T, S) B(T ) B(S) E

1 B(0) B(T ) XT 1 InfoT P (T, S) B(T ) B(S)

1 B(0) InfoT XT P (T, S) B(S)

1 B(0) XT P (T, S) B(S)

where X is a T -measurable random variable, known from InfoT

To value the above contract, notice that

1 B(0) B(Ti) P (Ti, Ti+1)

B(0) 1 P (Ti ,Ti+1 ) B(Ti+1 )

1 B(0) =E P (Ti, Ti+1)2 B(Ti+1)

8 9  = <X B(0) 1 B(0) =E (1 + i+1K) :i=+1 B(Ti+1) P (Ti, Ti+1)2 B(Ti) ;

Universit` Bocconi: Fixed Income a

1 B(0) (1 + i+1K) B(Ti+1) P (Ti, Ti+1)2 B(Ti)

B(0) (1 + i+1K) B(Ti)

P (0, Ti)(1 + i+1K)

P (0, Ti)(1 + i+1K).

Universit` Bocconi: Fixed Income a

dFi+1(t) = 2Fi+1(t)dFi+1(t) + 1 2dFi+1(t)dFi+1(t) 2 = i+1(t) Fi+1(t)dt + 2i+1(t)Fi+1(t)dZi+1(t) ,

dFi+1(t) = i+1(t) Fi+1(t)dt + 2i+1(t)Fi+1(t)dZi+1(t) ,

Universit` Bocconi: Fixed Income a

By expanding the square and substituting we obtain IAS

{P (0, Ti+1) [1 + 2i+1Fi+1(0)+

2 2 2 i+1Fi+1(0) exp(vi Ti)

(1 + i+1K)P (0, Ti)}.

Universit` Bocconi: Fixed Income a

Ratchet Caps and Floors

D(0, Ti)i [L(Ti1, Ti) (L(Ti2, Ti1) + X)] ,

X is a margin, which can be either positive or negative.

D(0, Ti)i [L(Ti1, Ti) Xi] ,

Xi = max L(Ti2, Ti1) X, Xi1 X , X := L(T1, T).

Ratchet caps and oors

D(0, Ti)i [L(Ti1, Ti) Xi] }

[L(Ti1, Ti) Xi] P (Ti, T )

Universit` Bocconi: Fixed Income a

Non-Sticky Ratchets: Analytical approximation

All we need to compute is the expectation

(t)Fi1(t)dt + i1(t)Fi1(t)dZi1(t), i1,iii(t)Fi(0) . 1 + iFi(0)

Universit` Bocconi: Fixed Income a

Non-Sticky Ratchets: Analytical approximation. Spread option

H = (awS1(T ) + bwS2(T ) wK) ,

8 9 = <X B(0) 1 B(0) =E (1 + i+1K) :i=+1 B(Ti+1) P (Ti, Ti+1)2 B(Ti) ;

2 X ij i,j i(t)j (t)Fi(t)Fj (t)