Materials For The Course: Here

Materials for the course
The course is based on book “Mathematical Modeling and Computation

in Finance: With Exercises and Python and MATLAB Computer Codes”,
by C.W. Oosterlee and L.A. Grzelak, World Scientific Publishing Europe
Ltd, 2019. For more details go here.
I Youtube Channel with courses can be found here.

I Slides and the codes can be found here.
Lech A. Grzelak Financial Engineering- Interest Rates and xVA 1 / 43
List of content
7.1. Pricing of Caplets/Floorlets

7.2. Pricing of Interest Rate Swaps
7.3. Pricing of Swaptions under the Black-Scholes Model
7.4. Jamshidian’s Trick
7.5. Swaptions under the Hull-White Model
7.6. Negative Interest Rates
7.7. Shifted Lognormal, Shifted Implied Volatility
7.8. Summary of the Lecture + Homework

Pricing of Caplets/Floorlets
Caplets/Floorlets
I The price of a caplet, with a strike price K , is given by:

1
V CPL (t0 ) = Nτk EQ max (`k (Tk−1 ) − K , 0) F(t0 )

M(Tk )
h i
= Nτk P(t0 , Tk )ETk max (`k (Tk−1 ) − K , 0) F(t0 ) .

I Now, assuming that the libor rate follows a lognormal distribution,

i.e.:
d`(t; Tk−1 , Tk ) = σk `(t; Tk−1 , Tk )dW k (t).
I If we consider Black’s model dynamics to value this option, then the
value of caplet k is given by:
Capletk (t0 ) = Nk τk P(t0 , Tk ) [`(t0 ; Tk−1 , Tk )N(d1 ) − Kk N(d2 )] , with

`(t0 ;Tk−1 ,Tk )
log K
+ 12 σk2 (Tk − t0 )
d1 = p ,
σk (Tk − t0 )
p
d2 = d1 − σk Tk − t0 .
Caplets/Floorlets
I The price of a caplet, with a strike price K , is given by:

1

V CPL (t0 ) = Nτk EQ max (`k (Tk−1 ) − K , 0) F(t0 )

M(Tk )
h i
= Nτk P(t0 , Tk )ETk max (`k (Tk−1 ) − K , 0) F(t0 ) .

I By the definition of the Libor rate, the (scaled) caplet valuation formula
can be written as,
V CPL (t0 )

1 1
= Nτk ETk max − 1 − K , 0 F(t0 )

P(t0 , Tk ) τk P(Tk−1 , Tk )
h i
Tk −Ār (τk )−B̄r (τk )r (Tk−1 )
= N ·E max e − 1 − τk K , 0 F(t0 )

h i
= N · e−Ār (τk ) ETk max e−B̄r (τk )r (Tk−1 ) − K̂ , 0 F(t0 ) ,

with K̂ = (1 + τk K )eĀr (τk ) .

I Alternative, based on the tower property, derivations can be found in the
book on page 381.
Caplets/Floorlets
I From Lecture 5 we know that a European-style option is defined by

the following equation:

M(t0 )
V ZCB (t0 , Tk−1 ) = EQ max (ᾱ(P(Tk−1 , Tk ) − K ), 0) F(t0 ) ,

M(Tk−1 )
with ᾱ = 1 for a call and ᾱ = −1 for a put option, strike price K

and dM(t) = r (t)M(t)dt.
I The price of a caplet with a strike price K can be written as,

1

V CPL (t0 ) = Nτi EQ EQ max (ì (Ti−1 ) − K , 0) F(Ti−1 )

M(Ti )

1 M(Ti−1 )

= Nτi EQ EQ max (ì (Ti−1 ) − K , 0) F(Ti−1 )

M(Ti−1 ) M(Ti )

Caplets/Floorlets
I After a change of measure, the inner expectation can be expressed
as,

M(Ti−1 )

EQ max (ì (Ti−1 ) − K , 0) F(Ti−1 ) = P(Ti−1 , Ti ) max (ì (Ti−1 ) − K , 0

M(Ti )
I The caplet value can therefore be found as,

1

V CPL (t0 ) = Nτi EQ P(Ti−1 , Ti ) max (ì (Ti−1 ) − K , 0) F(t0 ) .

M(Ti−1 )
I For the (scaled) caplet value, this results in
V CPL (t0 )

Ti−1 1 1
= Nτi · E P(Ti−1 , Ti ) max − 1 − K, 0

P(t0 , Ti−1 ) τi P(Ti−1 , Ti )

1
= N̂ · ETi−1 max − P(Ti−1 , Ti ) F(t0 ) ,

K̂
with N̂ = N (1 + τi K ) and K̂ = 1 + τi K .

Implied Volatility for Caplets/Floorlets
Effect of on implied volatility Effect of on implied volatility

22 45
=0.01 =0.01
=0.03 40 =0.02
20
=0.05 =0.03
=0.09 =0.04
35
18
implied volatility [%]
implied volatility [%]

30
16
25
14
20
12
15
10
10
8 5
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
K K
Figure: Effect of λ and η in the Hull-White model on the caplet implied

volatilities.

Implied Volatility for Caplets/Floorlets
I The impact of the different Hull-White model parameters, λ, η, on

the caplet implied volatility is presented.
I The mean reversion parameter λ appears to have a much smaller
effect on the implied volatilities than the volatility parameter η.
I In practice, therefore, λ is often kept fixed whereas η is determined
in a calibration process.
I Often the Hull-White model is extended and time-dependent
volatility parameter η(t) is used.
I The Hull-White model with time-dependent volatility parameter is
the market practice for xVA. There the calibration is performed to
(Bermudan) Swaptions.

Pricing of Interest Rate Swaps
Swaps
I Previously we have derived the value of a swap:
n
X n
X
Swap
Vm,n (t0 ) = τk P(t0 , Tk )`(t0 , Tk−1 , Tk ) − K τk P(t0 , Tk ).
k=m+1 k=m+1
I The first summation can be further simplified, using the definition of the
libor rate,
n n
X X P(t0 , Tk−1 ) − P(t0 , Tk )
τk P(t0 , Tk )`k (t0 ) = τk P(t0 , Tk )
τk P(t0 , Tk )
k=m+1 k=m+1
n
X
= P(t0 , Tk−1 ) − P(t0 , Tk )
k=m+1
= P(t0 , Tm ) − P(t0 , Tn ),
where in the last step the telescopic summation was recognized.
I The price of the swap is as follows:
n
X
Swap
Vm,n (t0 ) = [P(t0 , Tm ) − P(t0 , Tn )] − K τk P(t0 , Tk ).
k=m+1

Pricing of Interest Rate Swaps
Swaps
I By setting the value of the swap to zero, entering into such a deal is
for free. Moreover, the strike value for which the swap equals zero is
called swap rate and is indicated by Sm,n (t0 ).
I By equating the swap value to zero we find,
P(t0 , Tm ) − P(t0 , Tn ) P(t0 , Tm ) − P(t0 , Tn )
Sm,n (t0 ) = Pn = ,
k=m+1 τk P(t0 , Tk ) Am,n (t0 )
which, alternatively, by not simplifying the first summation in (1) we can

write,
n n
1 X X
Sm,n (t0 ) = τk P(t0 , Tk )`k (t0 ) = ωk (t0 )`k (t0 ),
Am,n (t0 )
k=m+1 k=m+1
with ωk (t0 ) = τk P(t0 , Tk )Am,n (t0 ) and `k (t0 ) := `(t0 , Tk−1 , Tk ).

I We can express the value of the swap as:
Swap
Vm,n (t0 ) = Am,n (t0 ) Sm,n (t0 ) − K .

Pricing of Swaptions under the Black-Scholes Model
European Swaptions
I Swaptions are the options on interest rate swaps, i.e. a holder of a

European swaption has the right, but not an obligation, to enter a
swap at a future date at a given predetermined strike K .
I Similarly as for swaps the swaptions have payers and receivers who
pay the fixed rate and receive the float leg (payer) or vice versa
(receiver).
I In the standard setting the swaption maturity coincides with the first
reset date of the underlying interest swap, i.e.: T0 = Tm .
I Let time t0 be today’s date and Tm be some future date at which
one has the option to enter into a swap deal, with the first reset
date given by Tm .

European Swaptions
I The value of the deal at time Tm is then given by:

S Swap
Vm,n (Tm ) = max Vm,n (Tm ), 0
" n #
X
= max τk P(Tm , Tk ) `(Tm , Tk−1 , Tk ) − K , 0 ,
k=m+1
and today’s discounted value is equal to:

" n
!#
S M(t0 ) X
Vm,n (Tm ) = EQ max τk P(Tm , Tk ) `(Tm , Tk−1 , Tk ) − K , 0 .
M(Tm )
k=m+1
I Using the annuity representation the value of the swaption is also equal to:

S M(t0 )
Vm,n (t0 ) = EQ max Am,n (Tm ) Sm,n (Tm ) − K , 0
M(Tm )

Q Am,n (Tm )M(t0 )
= E max (Sm,n (Tm ) − K , 0) ,
M(Tm )
where M(t0 ) = 1 and where t0 < Tm and Tm < Tm+1 , with Tm+1 the first
payment date.
European Swaptions
I As discussed, annuity Am,n (Tm ) is simply a combination of tradable
zero-coupon bonds and can therefore be considered as a numéraire.
I This suggests that we can define a Radon-Nikodym derivative for
changing measures, from the risk-neutral measure, Q, associated
with the money-savings account, M(t), to the new annuity measure
(also known as the swap measure), Qm,n , associated with the
annuity Am,n (t), i.e.,
dQm,n Am,n (Tm ) M(t0 )

λm,n
Q (Tm ) = = m,n .
dQ F (Tm ) A (t0 ) M(Tm )

I The value of the swaption now is given by:
Am,n (Tm )M(t0 ) Am,n (t0 ) M(Tm )

S
Vm,n (t0 ) = Em,n max (S m,n (Tm ) − K , 0) .
M(Tm ) Am,n (Tm ) M(t0 )
I After simplifications we find:
S
Vm,n (t0 ) = Am,n (t0 )Em,n [max (Sm,n (Tm ) − K , 0)] .

European Swaptions under the Black-Scholes Model
I To avoid arbitrage the swap rate Sm,n (Tm ), i.e.
P(t, Tm ) − P(t, Tn )
Sm,n (t) = ,
Am,n (t)
has to be a martingale under the swap measure associated with

annuity Am,n (t).
I This implies that the dynamics of the swap rate Sm,n (t) under the
swap measure, Qm,n , have to be driftless.
I A standard approach to price a swaption is via the Black-Scholes
model, i.e, assuming GBM process for Sm,n (t) under the annuity
measure Am,n (t) with the dynamics given by:
dSm,n (t) = σm,n Sm,n (t)dW m,n (t).
I Note that the lognormal assumptions implies only positive swap

rates!
European Swaptions under the Black-Scholes Model
I Then, the price of European Payer Swaption is given by:

S
Vm,n (t0 ) = Am,n (t0 )Em,n [max (Sm,n (Tm ) − K , 0)]
= Am,n (t0 ) Sm,n (t0 )FN (0,1) (d1 ) − KFN (0,1) (d2 ) .

I Then, the price of European Receiver Swaption is given by:

S
Vm,n (t0 ) = Am,n (t0 )Em,n [max (K − Sm,n (Tm ), 0)]
Am,n (t0 ) KFN (0,1) (−d2 ) − Sm,n (t0 )FN (0,1) (−d1 ) .

=
I In the pricing two constants d1 and d2 are given by:

S (t )
log m,nK 0 + 12 σ 2 Tm p
d1 = √ , d2 = d1 − σm,n Tm .
σm,n Tm

European Swaptions under the Hull-White model

I We need a slightly different representation for pricing with the
Hull-White model.
n
X n
X
τk P(Tm , Tk ) `k (Tm ) − K = 1 − P(Tm , Tn ) − K τk P(Tm , Tk )
k=m+1 k=m+1
n
X
= 1− ck P(Tm , Tk ),
k=m+1
with ck = K τk for k = m + 1, . . . , n − 1 and cn = 1 + K τn .

I This gives us the following expression for the swaption price:
" n
!#
S Tm
X
Vm,n (t0 ) = NP(t0 , Tm )E max 1 − ck P(Tm , Tk ), 0 .
k=m+1
I Now we will use the definition of the ZCB under the Hull-White model.
I The pricing above is not easy as the distribution of the sum of P(Tm , Tk )
is unknown.
Jamshidian’s Trick
Jamshidian’s Trick- Introduction

I Calulations of an expectation of the sum, is, due to linearity of
expectation, a simple task:
n
X n
X
E ψk (X ) = E[ψk (X )].
k=1 k=1
I The same principle holds when dealing with min / max or any other
function:
n
X n
X
E max(ψk (X ) − K , 0) = E[max(ψk (X ) − K , 0)].
k=1 k=1
I Unfortunately, this principle cannot be used when we deal with a

max / min of a sum, i.e.,
n
X n
X
E max ψk (X ) − K , 0 6= E max ψ(X )k − K , 0
k=1 k=1
I Computation of the LHS is not trivial! F.Jamshidian has found an

algorithm that allows us to compute the LHS, almost analytically.
I In 1989 F.Jamshidian presented an approach for transforming the

problem of calculating of maximum of a sum into a sum of certain
maximums, i.e.: for
n
X
A = max ψk (X ) − K , 0 ,
k=1
where ψk (X ) is a monotone increasing or monotone decreasing

sequence of functions ψk (X ) : R → R+ .
I Since each ψk (X ) is monotone in X the sum nk=1 ψk (X ) is as well
P
this means that there exists X ∗ such that
n
X
ψk (X ∗ ) − K = 0,
k=1
where X ∗ needs typically be determined with some search algorithm

like a Newton-Raphson discussed during previous lectures.

I Since
n
X n
X
ψk (X ∗ ) − K = 0, thus K = ψk (X ∗ ).
k=1 k=1
I Now the maximum becomes:
n n
! n
!
X X ∗
X ∗
A = max ψk (X ) − ψk (X ), 0 = max (ψk (X ) − ψk (X )) , 0 .
k=1 k=1 k=1
I The expression for A becomes:

n
! n
X ∗
X
A = max (ψk (X ) − ψk (X )) , 0 = (ψk (X ) − ψk (X ∗ )) 1X >X ∗ ,
k=1 k=1
which finally can be expressed as:

n
X
A= max (ψk (X ) − ψk (X ∗ ), 0) .
k=1
I This results is essenctal when pricing of Swaptions with the Hull-White

Model.
Jamshidian’s trick- Python Example
I Let us consider a sequence:
ψk (X ) = e−ti |X | , ti = {0, 1, 2, 3, 4, . . . , N},
I Normal distributed random variable X ∼ N (0, 1).

I Compute
" N
#
X
A = E max ψk (X ) − K , 0 ,
k=1
directly using Monte Carlo and by using Jamshidian’s trick. Plot

results against strike K .

Applicability of the trick
I In Lecture 3 we have discussed the HW2F model. As we see the

ZCBs/yield is not monotone therefore the Jamshidian’s trick cannot
be easily applied.
Figure: Dynamics of the yield curve for random market scenarios under the 1F
Hull-White model.

Swaptions under the Hull-White Model

I Having this results in mind we can come back to the pricing of
Swaptions under the Hull-White model.
I The Hull-White model is driven by the following dynamics:
dr (t) = λ (θ(t) − r (t)) dt + ηdW Q (t),

where function θ(t) is given explicitly in terms of the ZCBs,
Pmarket (t0 , T ), and where λ, η ∈ R.
I The zero-coupon bond P(t, T ) under the, affine, Hull-White process
r (t) can be expressed as
P(t, T ) = exp (A(t, T ) + B(t, T )r (t)) ,
I Functions A(t, T ) and B(t, T ) can be simply found when
calculating the following expectation
h RT i
P(t, T ) = EQ e− t r (s)ds F(t0 ) ,

with r (t) to be driven by the Hull-White process.

I The complete formulation of the ZCB under the Hull-White mode

reads
P(t, T ) = eA(t,T )+B(t,T )r (t) ,
I where functions A(t, T ) and B(t, T ) are, for τ = T − t, given by:
Z T
η2 −2λτ −λτ

A(t, T ) = − 3 + e − 4e − 2λτ + λ θ(z)B(z, T )dz,
4λ3 t
1
B(t, T ) = − 1 − e−λ(T −t) .
λ
I Ultimately, the formulation above can be further simplified and the
integration over θ(t) can be avoided. Then the ZCB will explicitly depend
on ZCBs from the market.
I We notice that for a given time t, P(t, Tk ) are monotone in r (t) for any
Tk > k.


I Now, we apply Jamishidian’s trick to pricing of swaptions under the
Hull-White Model.
" " n
##
S Q M(t0 ) X
Vm,n (t0 ) =E max τk P(Tm , Tk ) `(Tm , Tk−1 , Tk ) − K , 0 .
M(Tm )
k=m+1
I Since swaption expires at time TM , we take the ZCB P(t, Tm ) as the new
numéraire.
I After a change of measure, from the risk-free measure Q to the
Tm -forward measure QTm , gives,
" " n ##
X
S Tm
Vm,n (t0 ) = P(t0 , Tm )E max τk P(Tm , Tk ) `(Tm , Tk−1 , Tk ) − K , 0 .
k=m+1


I Recall that for `k (Tm ) := `(Tm , Tk−1 , Tk ) the following relation has
to hold:
n
X n
X
τk P(Tm , Tk ) `k (Tm ) − K = 1 − P(Tm , Tn ) − K τk P(Tm , Tk )
k=m+1 k=m+1
n
X
= 1− ck P(Tm , Tk ),
k=m+1
with ci = K τk for i = m + 1, . . . , n − 1 and cn = 1 + K τn .

I This gives us the following expression for the swaption price:
" n
!#
S Tm
X
Vm,n (t0 ) = NP(t0 , Tm )E max 1 − ck P(Tm , Tk ), 0 .
k=m+1
I Now we will use the definition of the ZCB under the Hull-White model.

Zero-Coupon Bond under the Hull-White model

I The pricing of Swaptions reads:
" n
!#
S Tm
X A(Tm ,Tk )+B(Tm ,Tk )r (Tm )
Vm,n (t0 ) = NP(t0 , Tm )E max 1 − ck e ,0 .
k=m+1
I Now we apply Jamshidian’s trick we seek r ∗ such that

n
X ∗
1− ck eA(Tm ,Tk )+B(Tm ,Tk )r = 0.
k=m+1
I Now we can substitute this result for 1 in the pricing equation:

" n n
!#
A+Br ∗
S Tm
X X A+Br (Tm )
Vm,n (t0 ) = NP(t0 , Tm )E max ck e − ck e ,0
k=m+1 k=m+1
" n
!#
A+Br ∗
X
Tm A+Br (Tm )
= NP(t0 , Tm )E max ck e −e ,0 .
k=m+1


I Now, using the Jamshidian’s method we switch between max of sum
to a sum of max:
" n
!#
A+Br ∗
X
S Tm A+Br (Tm )
Vm,n (t0 ) = NP(t0 , Tm )E max ck e −e ,0
k=m+1
n
∗
X h i
= NP(t0 , Tm ) ck ETm max eA+Br − eA+Br (Tm ) , 0
k=m+1
I Commonly, it is rewritten for a strike K̂ as

n
X h i
S
Vm,n (t0 ) = NP(t0 , Tm ) ck ETm max K̂ − eA+Br (Tm ) , 0 ,
k=m+1
with ∗
K̂ = eA+Br ,
∗
and where a constant r chosen such that:
n
X
1− ck exp A(Tm , Tk ) + B(Tm , Tk )r ∗ = 0.
k=m+1

I The task of pricing swaption is not complete yet as one still needs to
determine the sum of expectations.
I We notice that each element of this sum is simply an European put
option on the zero-coupon bond, i.e.:
h i
VTZCB
m ,Tk
(t0 , K , −1) = P(t0 , Tm )ETm max K − eAr (τk )+Br (τk )r (Tm ) , 0 ,
with τk = Tk − Tm .
I Pricing of European-type options on the ZCB can be, under the
Hull-White model, done analytically with the closed-form pricing solution.
I Closed form solution for pricing an option on a ZCB under the HW model
is discussed in Lecture 4.

I The final pricing equation for European swaptions then reads:

n
X
Swaption
Vm,n (t0 ) = N ck VTZCB
m ,Tk
(t0 , K̂ , −1),
k=m+1
with strike
K̂ := exp (Ar (Tm , Tk ) + Br (Tm , Tk )r ∗ )
where a constant r ∗ is pre-calibrated by solving of the following

equation:
n
X
1− ck exp Ar (Tm , Tk ) + Br (Tm , Tk )r ∗ = 0.
k=m+1

Negative Interest Rates
The Motivation Behind Negative Rates
I The financial crisis, which started 2007, exposed a lack of trust

among counterparties.
I Since then incorporation of the probability of default in pricing
models became standard.
I The lack of trust in the financial system reduced trading activities
and kept the money in the pockets.
I To recover trust central banks decided to intervene and stimulate
the monetary supply and demand.
I Lowering the interest rates was expected to encourage investors to
borrow money at a low rate and invest into the economy, which
would push the economy to grow.
I Since 2008 interest rates have gradually been lowered.

The Motivation Behind Negative Rates

I June 5th 2014, the rates set by ECB were negative for the first time:
minus 10 basis points.
I Using negative rates is unconventional to “inspire” investors but it is
not unprecedented (Switzerland, Sweden and Denmark also report
negative rates).
Interest Rates, EUR1M, 2008 Interest Rates, EUR1M, 2017
4.7 0.8
0.6
4.6
0.4
4.5
0.2
yield
yield
4.4 0
-0.2
4.3
-0.4
4.2
-0.6
4.1 -0.8
0 5 10 15 20 25 30 0 5 10 15 20 25 30
time in years time in years
Figure: Both figures present the yield obtained from EUR 1M curve, left: 2008
and right: 2017.
Modeling of Volatility
I Each non-linear product relies on volatilities that are extracted from
market quotes.
I Because of limited liquidity the volatilities are parameterized.
I Unfortunately, the parameterizations are often not arbitrage-free,
especially not in the low/negative interest rate environment (this will
be handled in a follow-up course).
Hagans implied PDF Hagan vs. the collocation implied density
20 20
Hagan density
15
Collocation Density
15
10
10
5
fY(y)
fY(y)
0 5
−5
0
−10
−5
−15
−20 −10
0 0.05 0.1 0.15 −0.15 −0.1 −0.05 0 0.05 0.1 0.15
y x
Figure: Probability density, with deterioration near zero; right: application of

the collocation method
Modeling Challenges and Client Relations

I Often, floating rate bonds are issued with coupons paid with a
spread above the index.
I Shall the bond issuer request the payments from bond holders in the
case the coupon payment is negative:
cpnk = L(Tk−1 , Tk−1 , Tk ) + spread < 0?
I To maintain a good client relation often coupons are calculated by
cpnk = max(L(Tk−1 , Tk−1 , Tk ) + spread, 0).
I This however requires incorporation of volatilities
F+ spr F+spr F+spr F+spr F+spr

F+ spr F+spr F+spr F+spr F+spr

Pricing of Caplets under negative rates

I An obvious example for which the pricing under negative rates needs
to be modified is the valuation of caplets.
I Based on the assumption of lognormal Libor rates,
`k (t) := `(t; Tk−1 , Tk ), with the corresponding dynamics,
d`k (t) = σk `k (t)dWkk (t),
the pricing equation of a caplet is given by:
h i
VkCPL (t0 ) = Nk τk P(t0 , Tk )ETk max (`k (Tk−1 ) − K , 0) F(t0 ) .

I The solution is given by

VkCPL (t0 ) = Nk τk P(t0 , Tk ) [`k (t0 )N(d1 ) − Kk N(d2 )] ,
with,

`k (t0 )
log K + 12 σk2 (Tk − t0 ) p
d1 = √ , d2 = d1 − σk Tk − t0 ,
σk Tk − t0
with `k (t0 ) = `(t0 ; Tk−1 , Tk ).
Shifted Lognormal, Shifted Implied Volatility
I In a negative interest rate environment, the ZCBs P(t0 , Tk ) may get

values that are higher than one unit of currency. This would imply
that the Libor rate `(t0 ; Tk−1 , Tk ) would be negative. When the
Libor rate is negative, the above pricing equation is not valid
anymore, as the logarithm of a negative value is not well-defined.
I Instead of GBM dynamics, the arithmetic Brownian motion (ABM)
dynamics, that also give rise to negative realizations could, for
example, be chosen. Such a solution, although straight-forward, has
a significant disadvantage, however, as the normal distribution has
much flatter distribution tails as compared to a lognormal process.
I Instead of completely changing the underlying dynamics, the
industrial standard for dealing with the negativity has become to
“shift” the original process.

I Caplets can be priced under a negative interest rate environment by

an adaptation of the underlying dynamics of the Libor rate.
I This shifted process is defined, as follows,
`ˆk (t) = `k (t) + θk ,
where the process `ˆk (t) is governed by a lognormal process, with the
following dynamics,
ˆ Tk−1 , Tk ) = σ̂k `(t;

d`(t; ˆ Tk−1 , Tk )dWkk (t).
I The concept of shifting a distribution is very similar to the idea of

“displacement”.

I The Libor rate `k (t), as observed in the market, is simply given by

`k (t) := `ˆk (t) − θk .
Shifting of the processes is equivalent to “moving” the probability
density along the x−axis.
I The pricing of caplets is now given by,
h i
VkCPL (t0 ) = Nk τk P(t0 , Tk )ETk max (`k (Tk−1 ) − K , 0) F(t0 )

h i
= Nk τk P(t0 , Tk )ETk max `ˆk (Tk−1 ) − θk − K , 0 F(t0 )

h i
= Nk τk P(t0 , Tk )ETk max `ˆk (Tk−1 ) − K̂ , 0 F(t0 ) ,

with K̂ = K + θk .

I Option pricing under shifted distributions is very convenient. The

corresponding solution is in accordance with the unshifted variant,
h i
VkCPL (t0 ) = Nk τk P(t0 , Tk ) `ˆk (t0 )N(d1 ) − K̂k N(d2 ) ,
with,
ˆ
`k (t0 )
log K̂
+ 12 σk2 (Tk − t0 ) p
d1 = √ , d2 = d1 − σk Tk − t0 ,
σk Tk − t0
with K̂ = K + θk and `ˆk (t0 ) = `k (t0 ) + θk .

Negative interest rates

I For decades the pricing models which generated negative rates were
considered unrealistic and simplistic.
I The Hull-White model, which “suffered” from negative rates, has
been widely used in the industry.
Hull-White, positive IR environment Hull-White, negative IR environment
30 15
20 10
density
density
10 5
0 0
0.1
0.1
1 0.05 1
0.05 0.8 0 0.8
0.6 0.6
-0.05
0 0.4 0.4
0.2 -0.1 0.2
r(t) r(t)
0 t 0 t
Figure: Both figures present the Monte Carlo Paths for the Hull-White model in
a positive and negative interest rate environment.

Negative interest rates
Shifted Lognormal
5
shift = 0
4.5 shift = 0.25
shift = 0.75
4 1.5
3.5
1
density
3
PDF
2.5 0.5
2
0
1.5 3
2
1
2
1
1.5
0.5 0
1
-1
0 0.5
F(t)
-1 -0.5 0 0.5 1 1.5 2 -2 0 t
x
Figure: Shifted lognormal distribution used for pricing in negative interest rate
environment.

Pricing under shifted distributions
Figure: Shifted lognormal distribution used for pricing in negative interest rate
environment.

Summary of the Lecture + Homework
Summary
I Pricing of Caplets/Floorlets
I Pricing of Interest Rate Swaps
I Pricing of Swaptions under the Black-Scholes Model
I Jamshidian’s Trick
I Swaptions under the Hull-White Model
I Negative Interest Rates
I Shifted Lognormal, Shifted Implied Volatility
I Summary of the Lecture + Homework

Summary of the Lecture + Homework
Homework Exercises
I Exercise
I Extend the code presented in slide 41 (pricing of caplets under
shifted lognormal) and compute the corresponding implied
volatilities.
I In the lecture we have presented how to use Jamshidian’s trick.
Apply the same strategy for pricing of Swaptions (the code for
pricing of options on ZCBs is included in today’s materials).
Compare your results to Monte-Carlo.

Materials For The Course: Here

Uploaded by

Document Information

Original Description:

Original Title

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

Materials For The Course: Here

Uploaded by

Copyright:

Available Formats

Materials for the course

The course is based on book “Mathematical Modeling and Computation

I Youtube Channel with courses can be found here.

7.1. Pricing of Caplets/Floorlets

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 2 / 43

I Now, assuming that the libor rate follows a lognormal distribution,

with K̂ = (1 + τk K )eĀr (τk ) .

I From Lecture 5 we know that a European-style option is defined by

with ᾱ = 1 for a call and ᾱ = −1 for a put option, strike price K

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 5 / 43

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 6 / 43

Implied Volatility for Caplets/Floorlets

Effect of on implied volatility Effect of on implied volatility

implied volatility [%]

Figure: Effect of λ and η in the Hull-White model on the caplet implied

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 7 / 43

Implied Volatility for Caplets/Floorlets

I The impact of the different Hull-White model parameters, λ, η, on

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 8 / 43

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 9 / 43

which, alternatively, by not simplifying the first summation in (1) we can

with ωk (t0 ) = τk P(t0 , Tk )Am,n (t0 ) and `k (t0 ) := `(t0 , Tk−1 , Tk ).

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 10 / 43

I Swaptions are the options on interest rate swaps, i.e. a holder of a

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 11 / 43

and today’s discounted value is equal to:

dQm,n Am,n (Tm ) M(t0 )

I The value of the swaption now is given by:

Am,n (Tm )M(t0 ) Am,n (t0 ) M(Tm )

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 13 / 43

European Swaptions under the Black-Scholes Model

I To avoid arbitrage the swap rate Sm,n (Tm ), i.e.

has to be a martingale under the swap measure associated with

dSm,n (t) = σm,n Sm,n (t)dW m,n (t).

I Note that the lognormal assumptions implies only positive swap

European Swaptions under the Black-Scholes Model

I Then, the price of European Payer Swaption is given by:

I Then, the price of European Receiver Swaption is given by:

I In the pricing two constants d1 and d2 are given by:

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 15 / 43

European Swaptions under the Hull-White model

with ck = K τk for k = m + 1, . . . , n − 1 and cn = 1 + K τn .

Jamshidian’s Trick- Introduction

I Unfortunately, this principle cannot be used when we deal with a

I Computation of the LHS is not trivial! F.Jamshidian has found an

I In 1989 F.Jamshidian presented an approach for transforming the

where ψk (X ) is a monotone increasing or monotone decreasing

where X ∗ needs typically be determined with some search algorithm

European Swaptions under the Hull-White model

I The expression for A becomes:

which finally can be expressed as:

I This results is essenctal when pricing of Swaptions with the Hull-White

Jamshidian’s trick- Python Example

I Let us consider a sequence:

ψk (X ) = e−ti |X | , ti = {0, 1, 2, 3, 4, . . . , N},

I Normal distributed random variable X ∼ N (0, 1).

directly using Monte Carlo and by using Jamshidian’s trick. Plot

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 20 / 43

Applicability of the trick

I In Lecture 3 we have discussed the HW2F model. As we see the

Lech A. Grzelak Financial Engineering- Interest Rates and xVA 21 / 43

European Swaptions under the Hull-White model