Quantitative Finance

Module 4 Exam

Bruce Haydon
May 21, 2015

Page <1>

Bruce Haydon

Question 1 YTM, Bond Convexity & Duration

P=$100
T = 3 yrs
V= Z(t,T)= $82

V =Pe y(Tt )
V
82
ln
P
100 1
yield maturity= y =
=
=
ln ( .82 )=0.066
T t
3
3
ln

( ) ( )

dV
=( T t ) P e y (T t )
dy

Duration=

1 dV 1
= ( T t ) P e y (Tt )
V dy V

1 d2 V 1
Convexity=
= (T t )2 P e y(T t ) =
2
V dy V

1
( 3 ) x 100 x e3 y = 3.00
82
1
82

32 x 100 x e3 y = 9.00

P = $100
T = 5 yrs
V = $90
C = 3% annually
N

V =Pe y(Tt )+ Ci e y(t t )

i=1

y5

92=100 x e

+ 3 x e

y(i)

i=1

Page <2>

Bruce Haydon

At this point we must use Excel Solver to iteratively sove for y as Yield To
Maturity of the coupon bond.

From using Excel Solver, we find YTM = 5.19%

1 dV
V dy

The duration is given by

dV
y ( T t )
y(t t )
=( T t ) P e
Ci e
dy
i=1
i

y (5 )

5 x 100 x e

3 ie y(i)
i=1

Duration=

90

) = 4.699 (from Excel table above)

Convexity is defined as:

1 dV
V dy

= -

(T t)

(5)

90

Page <3>

Bruce Haydon

 22.909

(from Excel table above)

M4Exam_Question1_
V1.xlsx

Page <4>

Bruce Haydon

Let r be an Ito process:

dr= A (r , t) dt+ B(r , t)d X

f
=exp ( X )=r
X
2

f
=exp ( X )=r
2
X

f = X

f 1 2 2 2 f
f
+ X
dt+ X
dX
2
X 2

X
X

1
df =d (e ln (r ))=dr= Xr+ 2 X 2 r dt+ ( Xr ) dX
2

1
rX + 2 X dt+ ( Xr ) dX
2

1
rlnr + 2 ln ( r ) dt + ( rln ( r ) ) dX
2

1
rlnr + 2 ln ( r ) dt+dX
2

((

Page <5>

Bruce Haydon

Page <6>

Bruce Haydon

The Extended Hull and White model takes the form

dr= ( ( t ) r ) dt +cdW
Thus, the pricing equation becomes:

V 1 2 V
+ c
+
t 2 r2

Lets assign

( t ) r

V
rV =0
r

V ( r ,t ; T ) as the value of a zero-coupon bond

Z ( r ,T ; T )

The final condition for a zero-coupon bond is:

A ( t ) rB(t)

Page <7>

Bruce Haydon

A= A ( t ; T )

B=B ( t ; T )

Now substitute this value for Z into the BPE, for which we need three
derivative terms:
Define

=tT

V ( r , t ; T )=e

A ( )rB ( )

define
V V
=
=( A ()r B( ) ) V
t

V
=B( ) V
r
2 V
2
=B( ) V
2
r
Substituting these expressions into the pricing formula for the
extended Hull & White model, and dividing through by V :

( A( )+ r B ( ) )+ 12 B( )2 c 2B( )

( )r r=0

1
B( )2 c 2B ( ) ( ) + B( ) r A ( )+ B ( ) rr =0
2
Grouping like terms together, we have an expression that is linear in
r:

1
B ( )2 c2B ( ) ( ) A () + B( ) + B ( )1 r =0
2

)(

Page <8>

Bruce Haydon

Accordingly, both of the expressions in parentheses must be zero. This

(1)

1
A ( )= c2 B ( )2B ( ) ( )
2

(2)

In order for the final condition at

t=T

to be satisfied, we need:

V ( r ,t ; T )=1
Hence

e A (T )rB (T )=0
A ( T )rB (T )=0
At

t=T , =0

This translates accordingly

A ( =0 )=B ( =0 )=0

As eq (2) only involves

B ( ) , it is possible to solve the two ODEs in a

recursive fashion. First we derive eq (1), and then by substitution back into
eq(2),

A ( )

can be found.

Re-writing (1):

Page <9>

Bruce Haydon

B ( )+ B ( )=1
Multiply both sides by factor

B ( ) e + B ( ) e =e
Examining the LHS, we recognize the structure as appearing to be the result
of the product rule for differentiation.

d
( e B ( ))=e
d
Use integration of both sides to obtain a general solution

d ( e B( ) )= e d
1
e B ( ) = e + c

1
B ( )= +c e

Going back and utilizing our final condition at

=0 that B(0) = 0

1
(0)
B ( 0 )= +c e

1
0= + c (1)

c=

Therefore, the particular solution of (1) satisfying the final condition B(0)=0
becomes

Page <10>

Bruce Haydon

1 1
B ( )= e

B ( )=

1e

(3)

Next is to consider a solution for

A ( )

which is obtained by integrating (2)

while incorporating the boundary condition that

A ( 0 )=0.

A ( )= A ( 0 )+ A ( s ) ds
0

1
A ( )= c2 B 2 ( s ) ds B ( s ) ( s ) ds
2 0
0

As such, we now need to find a solution for the integrals of

B 2 ( s)

and

B ( s) .

B ( s ) ( s ) ds=
0

1es
( s ) ds

1es
( s ) ds

2
And now for B ( s ) :

Page <11>

Bruce Haydon

s 2

) ds

1 2
1
1e
c B2 ( s ) ds= c 2
2 0
2 0

1
c2
2

( (
0

1 c2

2 2

1+e2 s +2 es
2

( ) (
( )[(
( )[(

e
2e
+
+
2

( )[(

1e
+
2

1 c

2 2

( ) [(

c 2

22

1+e2 s +2 es ) ds

e2 s 2 es

+
2

1 c

2 2

ds

1 c2

2 2

) )

)]

2 0

)(

)]

1 e2 2 e 2
+
+

2 2

)]

c 2
1 e2 2 e 4

2
2 2

2
2

( )(

c 2
e2 2 e 3

2
2

2
2

( )(

(e 1)
+2

)]

e
2e
0
+
2

Therefore, the particular solution satisfying the boundary conditions therefore

Page <12>

Bruce Haydon

B ( )=

A ( )= B ( s ) ( s ) ds+
0

Substituting back

c 2
e2 2 e 3

2
2

2
2

( )(

=T t

B ( t ;T )=

A ( )= B ( s ) ( s ) ds+
t

1e

( )(
c
2
2

1e (T t)

2 (Tt )

(T t)

(T t)

2e

3
2

and accordingly, bond prices are given as

V ( r ,t ; T )=e A ( )rB()

where

=T t

QED

Page <13>

Bruce Haydon

Upon inspection, this appears anon-mean-reverting variation of the OrnsteinUhlenbeck SDE.

d U t= U t dt +d X t

(4.1)

U 0=u 0

given

U ( s ) ds+ d X s
0

(4.2)

U t=u 0
0

Multiply both sides of (4.1) by an integrating factor

et

e d U t = e U t dt + e d X t

(4.3)

In addition, the chain rule tells us

d ( e t U t ) =e t d U t +U t d ( e t ) =e t d U t + U t e t dt
(4.4)

Substituting (4.3) into (4.4) yields the following

d ( e t U t ) =( et U t dt+ et d X t )+ U t e t dt

Page <14>

Bruce Haydon

d ( e t U t ) = e t d X t

Since

d ( e t U t ) = e t d X t , we can now integrate much easier as follows:

e U t u0= e s d X s
t

and so
t

U t =et u0 + e (t s) d X t

(4.5)

The integral in 4.5 is an Weiner integral, so by definition:

e
0

d X t N 0, e

2 ( t s)

) (

2 t

d s =N 0,

1
2

U 0=u 0 as a constant implies that

Choosing

e
( 2 t1)
2
t
e u0,
2

U t =et u0 + e (t s) d X t N
0

e
(2 t1)
and variance V[
2
U t =
2

Therefore

[ U t ] =mean=e

u0

Page <15>

Bruce Haydon

The bond pricing equation is:

V 1 2 2 V (
V
+ w
+ w )
rV =0
2
t 2
r
r
A final condition must be specified for this PDE. In the case of a zero-coupon bond,
the final condition is a time condition corresponding to the payoff at maturity T:

V ( r ,T ; T )=1
We need to do pricing for zero-coupon bonds in the risk-neutral world, where there
will be a number of parallel Martingale measures. Therefore, the actual pricing of
the bond will take place in the risk-neutral world Q .
Applying the fundamental asset pricing formula to a zero-coupon bond maturing at
time T, with value 1 at maturity, the value at time t (t T

r ( s) ds

B ( t ,T )=E Q e

F t

is given as:

This represents the Fundamental Asset Pricing Model (FAPM) applied to a zerocoupon bond. Pricing is about finding the expected present value of all cashflows
under a risk-neutral condition.
In other words, the zero coupon bond is a pure discount factor, and the price

B ( t ,T )
measure

represents the expectation under the pricing measure (risk-neutral

Q ) of all cash flows discounted back to time t.

The best way to obtain a price is to apply Feynman-Kac to the Fundamental Asset
Pricing Model (FAPM) to obtain a PDE, then find a solution to that PDE.
Given

Page <16>

Bruce Haydon

r ( s) ds

B ( t ,T )=E e

F t

and given that the rate under the risk neutral measure Q

is represented by

dr ( t ) =t dt + t d X (t)
we can use Feynman-Kac to obtain the pricing PDE associated with the expectation
above along with the

Q -process as follows:

2
B ( )
B
t , r + ( t t t )
( t ,r ( t ) ) + 1 2 ( t , r ) B2 ( t , r ( t )) r ( t ) B (t , r )=0
t
r
2
r

with terminal-value condition

B ( T , r (T ) )=1
The relationship between the physical measure

and an equivalent Martingale

measure is established by the market price of risk which acts as the change of

measure process. The market price of risk represents the compensation paid by the
market to an investor per unit of risk.

Page <17>

Bruce Haydon

Page <18>

Bruce Haydon