Stochastic Calculus for Finance II

46-945
Spring 2025

Solutions to Homework Set 5

Exercise 1 (Interest rate caplet in the Hull-White model). Consider the Hull-White
model of Homework 4. A caplet in this model pays (R(T ) − K)+ at time T .

(i) Compute the price of this caplet at time t for 0 ≤ t ≤ T . Write your answer in terms of
T
σ2
Z
2
e−2κ(T −u) du = 1 − e−2κ(T −t) , and d t, f (t, T ) ,



B(t, T ), v(t) = σ
t 2κ
where
x−K
d(t, x) = p
v(t)
and f (t, T ) is the forward interest rate at time t for borrowing at time T . Warning:
One can use the ideas behind Black’s formula here, but cannot apply Black’s formula
directly.

(ii) Define a function c(t, x) such
that the caplet price you computed in part (i) can be
written as B(t, T )c t, f (t, T ) .

Solution.

(i) According to Exercise 4(iv) of Homework 4, the forward interest rate in the Hull-White
model has differential
df (u, T ) = σe−κ(T −u) dW T (u),
where W T is the Brownian motion under the T -forward measure PT . Therefore,
Z T p
R(T ) = f (T, T ) = f (t, T ) + σ e−κ(T −u) dW T (u) = f (t, T ) − v(t) Z,
t

where
T
σ2
Z
2
e−2κ(T −u) du = 1 − e−2κ(T −t)


v(t) = σ
t 2κ
RT −κ(T −u)
is the variance of σ t
e dW T (u) and
Z T
σ
Z = −p e−κ(T −u) dW T (u)
v(t) t

1
 is a standard normal random variable under PT that is independent of F(t). The time-t
price of the caplet is
1 e
+ 
E D(T ) R(T ) − K F(t)
D(t)
 
B(0, T ) e D(T )B(T, T )
+
= B(t, T ) E R(T ) − K F(t)
D(t)B(t, T ) B(0, T )
h p
+ i
= B(t, T )ET f (t, T ) − v(t) Z − K F(t)
p
+ 
= B(t, T )ET f (t, T ) − v(t) Z − K

" !+ #
p f (t, T ) − K
= B(t, T ) v(t) ET p −Z
v(t)
p
+ 
= B(t, T ) v(t) ET d(t, f (t, T )) − Z

,

where we have used the independence between Z and F(t) in the third equation. We
compute
Z d(t,x)

+  1
2
T
d(t, x) − z e−z /2 dz

E d(t, x) − Z =√
2π −∞
d(t, x) d(t,x) −z2 /2
Z Z d(t,x)
1 2
= √ e dz − √ ze−z /2 dz
2π −∞ 2π −∞
1 2 z=d(t.x)
= d(t, x)N d(t, x) + √ e−z /2


2π z=−∞
0



= d(t, x)N d(t, x) + N d(t, x) .

The caplet price is thus

p h
i
0



B(t, T ) v(t) d t, f (t, T ) N d(t, f (t, T )) + N d(t, f (t, T ))
p
= B(t, T ) f (t, T ) − K N d(t, f (t, T )) + B(t, T ) v(t) N 0 d(t, f (t, T )) .

(ii) We define

p
c(t, x) = (x − K)N d(t, x) + v(t) N 0 d(t, x) .

Exercise 2 (Hedging with futures in the Hull-White model). In Example 2.11 of

Lecture Notes for Fixed-Income Model, in the Ho-Lee model we computed the forward price
ForB (t, T1 ) at time t for delivery at time T1 of the T2 -maturity bond in the Ho-Lee model,
where 0 ≤ t ≤ T1 < T2 .
This exercise is similar to Example 2.11, except that we use the Hull-White model rather
than the Ho-Lee model in this exercise. The differential of the spot rate in the Hull-White
model is

dR(u) = κ θ(u) − R(u) du + σdW f (u),

2
where κ and σ are positive constants, θ(u) is a non-random function, and W
f is a Brownian
motion under a risk-neutral measure P.
e The prices of default-free zero-coupon bonds in this
model were shown in Exercise 2 of Homework 4 to be of the form

B(t, T ) = e−C(t,T )R(t)−A(t,T ) , 0 ≤ t ≤ T.

where C(t, T ) and A(t, T ) are nonrandom function computed in Exercise 2 of Homework 4.
Although we have formulas for C(t, T ) and A(t, T ), in this exercise and its solution we will
just use the notation C(t, T ) and A(t, T ) without writing out the formulas.

(i) In Homework 4, Exercise 4, you showed that the volatility of the T -maturity bond in
the Hull-White model is −σC(t, T ), 0 ≤ t ≤ T . The forward price for delivery at time
T1 of the T2 -maturity bond is

B(t, T2 )
ForB (t, T1 ) = .
B(t, T1 )

This is a quotient martingales. More precisely, it is the P-martingale

e D(t)B(t, T2 )
divided by the P-martingale D(t)B(t, T1 ). Use Exercise 4 of Homework 1 to derive the
e
formula for dForB (t, T1 ), first in terms of the T1 -forward Brownian motion W T1 and
then in terms of W
f.

(ii) Compute the price at time t ∈ [0, T1 ] of the call option that pays

+
B(T1 , T2 ) − K

and time T1 . Write the call price as

B(t, T1 )c t, F orB (t, T1 ) .

What is the function c(t, x)? Write it using the notation

 
1 x 1 2
d± (t, x) = √ log ± σ (t)(T1 − t) ,
σ(t) T1 − t K 2
s
Z T1
1
2
σ(t) = σ C(u, T1 ) − C(u, T2 ) du.
T1 − t t

We now set out to replicate the call option in part (ii) be trading interest rate futures
with futures price FutR (t, T1 ) computed in part (iii) below. This trading is financed using
the money market account. In particular, we want to begin with initial capital X(0) equal
to the initial price of the call B(0, T1 )c(0, ForB (0, T1 )) and hold Φ(t) futures contracts at
each time t so that X(t) = B(t, T1 )c(t, ForB (t, T1 )) for 0 ≤ t ≤ T1 . We begin by computing
the futures price and its differential.

3
(iii) Let 0 ≤ T1 < T2 be given. Compute
 
e R(T1 ) F(t) ,
FutR (t, T1 ) = E 0 ≤ t ≤ T1 ,

the futures price at time t for delivery of R(T1 ) at time T1 .

(iv) Compute dFutR (t, T1 ). This differential should have a dW

f term and no dt term because
futures prices are martingales under P.e

(v) In order to replicate the call option in part (ii) by trading interest rate futures, how
many futures contracts Φ(t) should be held at each time t? You may use without
verifying the formulas

σ2x
2
C(t, T1 ) − C(t, T2 ) N 0 d+ (t, x) ,


ct (t, x) = − √ (5.1)
2σ(t) T1 − t


cx (t, x) = N d+ (t, x) , (5.2)
1
N 0 d+ (t, x) ,


cxx (t, x) = √
σ(t)x T1 − t
1
2
0 = ct (t, x) + σ 2 C(t, T1 ) − C(t, T2 ) x2 cxx (t, x). (5.3)
2

Remark. In practice, one would use SOFR futures rather than spot rate futures as the
replicating (hedging) instrument for the bond call option of this exercise. The calculations
using SOFR futures are longer but not more instructive than the calculations using spot
rate futures, and so we use spot rate futures here. However, the formula obtained for Φ(t)
when we use spot rate futures has R(t) in the denominator, and in the Hull-White model,
R(t) can be zero. Thus, an attempt to implement the formula for Φ(t) obtained here might
result in Φ(t) blowing up.
Solution.

(i) Exercise 4 of Homework 1 tells us the following. The volatility of a quotient of mar-
tingales is the volatility of the numerator minus the volatility of the denominator.
Furthermore, the quotient is a martingale under the measure obtained using the de-
nominator, divided by its initial condition, as the Radon-Nikodym derivative process
to change the measure. In this case, the quotient of P-martingales
e is

D(t)B(t, T2 )
ForB (t, T1 ) = , 0 ≤ t ≤ T1 ,
D(t)B(t, T1 )

the Radon-Nikodym derivative process is

D(t)B(t, T1 )
, 0 ≤ t ≤ T1 ,
B(0, T1 )

4
 and the measure obtained when we change from P e using this Radon-Nikodym derivative
T1
process is the T1 -forward measure P . The Brownian motion under this measure is
Z t
T1
W (t) = Wf (t) + σ C(u, T1 )du
0

because the volatility process for the denominator of the quotient is −σC(t, T1 ). There-
fore,

dForB (t, T1 ) = σ C(t, T1 ) − C(t, T2 ) ForB (t, T1 )dW T1 (t)

= σ C(t, T1 ) − C(t, T2 ) ForB (t, T1 ) dW
f (t) + σC(t, T1 )dt .

(ii) We take T in Theorem 2.9 of Lecture Notes for Fixed-Income Models to be T1 and take
S(t) to be B(t, T2 ). According to Remark 2.10 in the lecture notes, the call price at
time t ∈ [0, T1 ] is



B(t, T2 )N d+ (t, ForB (t, T1 )) − KB(t, T1 )N d− (t, ForB (t, T1 ))
h

i
= B(t, T1 ) ForB (t, T1 )N d+ (t, ForB (t, T1 )) − KN d− (t, ForB (t, T1 )) .

The function c(t, x) is

c(t, x) = xN d+ (t, x) − KN d− (t, x) .

(iii) According to Homework 4, Exercise 1, for 0 ≤ t ≤ T1 , we have

Z T1 Z T1
−κ(T1 −t) −κ(T1 −u)
R(T1 ) = e R(t) + κ e θ(u)du + σ e−κ(T1 −u) dW
f (u).
t t

Therefore,
 Z T1 Z T1 
−κ(T1 −t) −κ(T1 −u) −κ(T1 −u) f (u) F(t)
FutR (t, T1 ) = E
e e R(t) + κ e θ(u)du + σ e dW
t t
Z T1
−κ(T1 −t)
=e R(t) + κ e−κ(T1 −u) θ(u)du.
t

(iv) We compute

dFutR (t, T1 ) = κe−κ(T1 −t) R(t)dt + e−κ(T1 −t) dR(t) − κe−κ(T1 −t) θ(t)dt
= κe−κ(T1 −t) R(t)dt + e−κ(T1 −t) κ(θ(t) − R(t) dt + σR(t)dW


f (t)
− κe−κ(T1 −t) θ(t)dt
= e−κ(T1 −t) σR(t) dW
f (t).

5
(v) The value of the portfolio that holds futures and finances the trading using the money
market account has differential
dX(t) = Φ(t)dFutB (t, T1 ) + R(t)X(t)dt.
The differential of the discounted portfolio value is


d D(t)X(t) = −R(t)D(t)X(t)dt + D(t)dX(t)
= D(t)Φ(t)dFutB (t, T1 )
= D(t)Φ(t)e−κ(T1 −t) σR(t)dW
f (t). (5.4)

We use the formula for dForB (t, T1 ) derived in part (i) and equations (5.3) and (5.2)
to compute


dc t, ForB (t, T )



= ct t, ForB (t, T1 ) dt + cx t, ForB (t, T1 ) dForB (t, T1 )
1

+ cxx t, ForB (t, T1 ) dForB (t, T1 )dForB (t, T1 )
 2 
1
2
2 2
= ct t, ForB (t, T1 ) + cxx t, ForB (t, T1 ) σ C(t, T1 ) − C(t, T2 ) ForB (t, T1 ) dt
2




+ cx t, ForB (t, T1 ) σ C(t, T1 ) − C(t, T2 ) ForB (t, T1 ) dW
f (t) + σC(t, T1 )dt




= N d+ (t, ForB (t, T1 )) σ C(t, T1 ) − C(t, T2 ) ForB (t, T1 ) dWf (t) + σC(t, T1 )dt .

The differential of the discounted call price is



d D(t)B(t, T1 )c t, ForB (t, T1 )




= d D(t)B(t, T1 ) · c t, ForB (t, T1 ) + D(t)B(t, T1 ) · dc t, ForB (t, T1 )



+ d D(t)B(t, T1 ) · dc t, ForB (t, T1 )


= −σC(t, T1 )D(t)B(t, T1 )c t, ForB (t, T1 ) dWf (t)



+ D(t)B(t, T1 )N d+ (t, ForB (t, T1 )) σ C(t, T1 ) − C(t, T2 ) ForB (t, T1 )


× dW f (t) + σC(t, T1 ) dt



− σC(t, T1 )D(t)B(t, T1 )N d+ (t, ForB (t, T1 )) σ C(t, T1 ) − C(t, T2 ) ForB (t, T1 )dt
h

= σD(t)B(t, T1 ) − C(t, T1 )c t, ForB (t, T1 )


i
+ N d+ (t, ForB (t, T1 )) C(t, T1 ) − C(t, T2 ) ForB (t, T1 ) dW f (t). (5.5)

Setting (5.4) and (5.5) equal, we see that

κ(T1 −t) B(t, T1 )

h

Φ(t) = e − C(t, T1 )c t, ForB (t, T1 )
R(t)


i
+ N d+ (t, ForB (t, T1 )) C(t, T1 ) − C(t, T2 ) ForB (t, T1 ) .

6
Exercise 3 (SOFR caplet in the Ho-Lee model). According to Definition 3.2 of the
Lectures Notes on Fixed-Income Models, the forward price for delivery at time Tj+1 of SOFR
set at time Tj+1 for the accrual period [Tj , Tj+1 ] is

B(t, Tj ) − B(t, Tj+1 )
, 0 ≤ t ≤ Tj ,


τ B(t, Tj+1 )

ForS (t; Tj , Tj+1 ) =
D(Tj ) 1
− , Tj ≤ t ≤ Tj+1 ,


τ D(t)B(t, Tj+1 ) τ


where τ = Tj+1 − Tj . In particular,

"Z # !
Tj+1
1
ForS (Tj+1 ; Tj , Tj+1 ) = S(Tj+1 ; Tj , Tj+1 ) = exp R(u)du − 1
τ Tj

is SOFR set at time Tj+1 for the accrual period [Tj , Tj+1 ].

(i) For the Ho-Lee model, compute the differential

 
1
d ForS (t; Tj , Tj+1 ) + for 0 ≤ t ≤ Tj+1 .
τ
Hint: For both the cases 0 ≤ t ≤ Tj and Tj ≤ t ≤ Tj+1 , the process ForS (t; Tj , Tj+1 ) +
1/τ can be written as a quotient of P-martingales.
e You computed the differential of a
quotient of martingles in Exercise 4 of Homework 1. Using the result of that exercise
and the Ho-Lee bond differential formula (1.52) from Lecture Notes on Fixed-Income
Models, you can write down the answer to this part of the problem without doing any
computation.
(ii) Compute "  + #
1
ETj+1 ForS (Tj+1 ; Tj , Tj+1 ) + −K .
τ
(For this part of the exericise, simplify notation by writing
1
X(t) = ForS (t; Tj , Tj+1 ) + , 0 ≤ t ≤ Tj+1 .
τ
The quantity  
2 2 1
v = σ τ Tj + τ
3
should appear in your answer.)
(iii) Determine the price at time zero of a caplet on SOFR, i.e., a call option that pays

+
S(Tj+1 ; Tj , Tj+1 ) − K

at time Tj+1 .

7
Solution.

(i) First observe that

B(t, Tj )


 , 0 ≤ t ≤ Tj ,
1 
τ B(t, Tj+1 )
ForS (t; Tj , Tj+1 ) + =
τ D(Tj )
, Tj ≤ t ≤ Tj+1 .


τ D(t)B(t, Tj+1 )


For 0 ≤ t ≤ Tj , this is the quotient of the P-martingales

e D(t)B(t, Tj ) and τ D(t)B(t, Tj ).
For Tj ≤ t ≤ Tj+1 , this is the quotient of the constant P-martingale
e D(Tj ) and the
P-martingale τ D(t)B(t, Tj+1 ). In both cases, the denominator divided by its initial
e
condition is the Radon-Nikodym derivative process (see Theorem 1.8 in Lecture Notes
on Fixed-Income Models)
Z t
1 2 t
 Z 
D(t)B(t, Tj+1 ) 2
= exp −σ (Tj+1 − u)dW (u) − σ
f (Tj+1 − u) du , 0 ≤ t ≤ Tj+1 ,
B(0, Tj+1 ) 0 2 0

that changes from the P e measure to the Tj+1 -forward measure PTj+1 . The Brownian
Tj+1
motion under P is
Z t
Tj+1
W (t) = W (t) + σ
f (Tj+1 − u)du, 0 ≤ t ≤ Tj+1 .
0

In the case 0 ≤ t ≤ Tj , the volatility of ForS (t; Tj , Tj+1 )+1/τ is the volatility −σ(Tj −t)
of D(t)B(t, Tj ) minus the volatility −σ(Tj+1 − t) of τ D(t)B(t, Tj+1 ), which is

−σ(Tj − t) + σ(Tj+1 − t) = σ(Tj+1 − Tj ) = στ.

In addition, ForS (t; Tj , Tj+1 ) + 1/τ is a martingale under PTj+1 . Therefore,

   
1 1
d ForS (t; Tj , Tj+1 ) + = στ ForS (t; Tj , Tj+1 ) + dW Tj+1 (t), 0 ≤ t ≤ Tj .
τ τ

In the case Tj ≤ t ≤ Tj+1 , the volatility of ForS (t; Tj , Tj+1 ) + 1/τ is the volatility 0 of
the constant D(Tj ) minus the volatility −σ(Tj+1 − t) of τ D(t)B(t, Tj+1 ), which is

σ(Tj+1 − t).

In addition, ForS (t; Tj , Tj+1 ) + 1/τ is a martingale under PTj+1 . Therefore,

   
1 1
d ForS (t; Tj , Tj+1 ) + = σ(Tj+1 −t) ForS (t; Tj , Tj+1 ) + dW Tj+1 (t), Tj ≤ t ≤ Tj+1 .
τ τ

8
(ii) To simplify notation, we define
1
X(t) = ForS (t; Tj , Tj+1 ) + .
τ
We showed in part (i) that
dX(t) = σ I[0,Tj ) (t)τ + I[Tj ,Tj+1 ] (t)(Tj+1 − t) X(t)dW Tj+1 (t),

which implies that

 Z Tj+1
I[0,Tj ) (t)τ + I[Tj ,Tj+1 ] (t)(Tj+1 − t) dW Tj+1 (t)


X(Tj+1 ) = X(0) exp σ
0
1 2 Tj+1
Z 

2
− σ I[0,Tj ) (t)τ + I[Tj ,Tj+1 ] (t)(Tj+1 − t) dt .
2 0

We note that
Z Tj+1  
2

2 2 2 1
v=σ I[0,Tj ) (t)τ + I[Tj ,Tj+1 ] (t)(Tj+1 − t) dt = σ τ Tj + τ
0 3
is the variance of
Z Tj+1
I[0,Tj ) (t)τ + I[Tj ,Tj+1 ] (t)(Tj+1 − t) dW Tj+1 (t).


σ
0

Therefore, we can write X(Tj+1 ) as

√


1
X(Tj+1 ) = X(0) exp vZ − v ,
2
where Z Tj+1
σ
I[0,Tj ) (t)τ + I[Tj ,Tj+1 ] (t)(Tj+1 − t) dW Tj+1 (t)


Z=√
v 0
is a standard normal random variable under PTj+1 . It is now a straightforward compu-
tation to show that
"  + #
1
ETj+1 ForS (Tj+1 ; Tj , Tj+1 ) + −K
τ

+ 
= ETj+1 X(Tj+1 ) − K


= X(0)N (d+ ) − KN (d− )

 
1
= ForS (0; Tj , Tj+1 ) + N (d+ ) − KN (d− ),
τ
where
 
1 X(0) 1
d± = √ log ± v
v K 2
  
1 ForS (0; Tj , Tj+1 ) + 1/τ 1 2 2 1
= q log ± σ τ Tj + τ .
στ Tj + 13 τ K 2 3

9
(iii) We use the fact that
   +


+ 1 1
S(Tj+1 ; Tj , Tj+1 − K = ForS (Tj+1 ; Tj , Tj+1 ) + − K+ .
τ τ
According to the risk-neutral pricing formula, the price at time zero of the caplet is
e D(Tj+1 ) S(Tj+1 ; Tj , Tj+1 ) − K +


E
 
D(T j+1 )B(T j+1 , Tj+1 )
+
= B(0, Tj+1 )E
e S(Tj+1 ; Tj , Tj+1 ) − K
B(0, Tj+1 )

+ 
= B(0, Tj+1 )ETj+1 S(Tj+1 ; Tj , Tj+1 ) − K

"   + #
1 1
= B(0, Tj+1 )ETj+1 ForS (Tj+1 ; Tj , Tj+1 ) + − K+ .
τ τ

This last expectation is obtained from the formulas in part (ii) when we replace K in
those formulas by K + 1/τ . We conclude that the time-zero price of the caplet is
    
1 1
B(0, Tj+1 ) ForS (0; Tj , Tj+1 + N (d+ ) − K + N (d− ) ,
τ τ
where
  
1 ForS (0; Tj , Tj+1 ) + 1/τ 1 2 2 1
d± = q log ± σ τ Tj + τ .
στ Tj + 13 τ K + 1/τ 2 3

Exercise 4. (Swaps measure). Because we did not have time in class to discuss
SOFR swaps, this exercise is optional. It will not be graded. To do it, you need
to first read Section 3.5 in the Lecture Notes on Fixed Income Models.
Recall Theorem 3.7 in the lecture notes, which says that the price of a payer swaption
beginning at time Ti with first payment at time Ti+1 and last payment at time Tk with
coupon rate c is
1 e
+
E D(Ti ) SRi,k (Ti ) − c Ai,k (Ti ) F(t) ,

Swaption(t) = 0 ≤ t ≤ Ti .
D(t)

In this formula, SRi,k (Ti ) is the swap rate given by (3.11) in the lecture notes,
1 − B(Ti , Tk )
SRi,k (Ti ) = ,
Ai,k (Ti )

and Ai,k (Ti ) is the annuity given by (3.9) in the lecture notes,
k−1
X
i,k
A (t) = τ B(t, Tj+1 ), 0 ≤ t ≤ Ti .
j=i

10
(i) Specify a Radon-Nikodym derivative process, defined for 0 ≤ t ≤ Ti , that will change
from the risk-neutral measure P e to a so-called swaps measure PSw such that

+
Swaption(t) = Ai,k (t)ESw SRi,k (Ti ) − c F(t) , 0 ≤ t ≤ Ti .
 

Explain how you know that the Radon-Nikodym derivative process you specify is a
positive martingale under P
e with initial condition equal to 1.

(ii) Show that the forward swap rate

B(t, Ti ) − B(t, Tk )
SRi,k (t) = , 0 ≤ t ≤ Ti ,
Ai,k (t)
given by equation (3.12) in the lecture notes, is a martingale under PSw .

Solution.

(i) We take the Radon-Nikodym derivative process to be

k−1
D(t)Ai,k (t) τ X
= D(t)B(t, Tj+1 ), 0 ≤ t ≤ Ti ,
Ai,k (0) Ai,k (0) j=i

This is a positive martingale under P e because D(t)B(t, Tj+1 ), 0 ≤ t ≤ Ti ≤ Tj+1 is a

positive martingale for each j = i, i + 1, . . . , k − 1. It is obviously equal to 1 if we set
t = 0. Finally, Bayes’ Rule implies
1 e
+
E D(Ti ) SRi,k (Ti ) − c Ai,k (Ti ) F(t)

Swaption(t) =
D(t)
Ai,k (0) e D(Ti )Ai,k (Ti )
 
i,k i,k

+
= A (t) · E SR (Ti ) − c F(t)
D(t)Ai,k (t) Ai,k (0)

+
= Ai,k (t)ESw SRi,k (Ti ) − c F(t) .
 

(ii) Let 0 ≤ s ≤ t ≤ Ti be given. According to Bayes’ Rule,

Ai,k (0) e D(t)Ai,k (t) i,k
 
Sw
 i,k 
E SR (t) F(s) = E SR (t) F(s)
D(s)Ai,k (s) Ai,k (0)
 
1 i,k B(t, Ti ) − B(t, Tk )
= E D(t)A (t)
e F(s)
D(s)Ai,k (s) Ai,k (t)
1 
e D(t)B(t, Ti ) − D(t)B(t, Tk ) F(s)

= E
D(s)Ai,k (s)
1

= D(s)B(s, T i ) − D(s)B(s, T k )
D(s)Ai,k (s)
B(s, Ti ) − B(s, Tk )
=
Ai,k (s)
= SRi,k (s).
This is the martingale property for the swap rate under PSw .

11