Вы находитесь на странице: 1из 8

CQF Module 3 Examination

Notes:
There are two sections.
Section A contains mathematical questions. You are required to present hand worked solutions.
Section B is a computational part. This requires some element of programming.
Answer all questions.
If you are unclear as to what a question requires, clarication should be sought from the tutor:
riaz.ahmad@tchlearning.com
Do not send solutions or partial answers for checking/comments.
The nal submission should contain clear neat working presented in an orderly fashion.
Questions which require the implementation of an algorithm using a programming language are clearly
indicated.
You are given the following
d
1
=
log (o,1) +
_
: +
1
2
o
2
_
(1 t)
o
_
1 t
, d
2
=
log (o,1) +
_
:
1
2
o
2
_
(1 t)
o
_
1 t
and
`(r) =
1
_
2:
_
x
1
exp(c
2
,2)dc . `
0
(r) =
1
_
2:
exp(r
2
,2)
1
Section A
1. 1 (t) = nA
1
(t) +
_
1 n
2
A
2
(t) is a model to construct a process from two uncorrelated Brownian
Motions. The processes A
1
(t). A
2
(t) have drifts in addition to the increment d\ of Wiener process:
dA
1
(t) = j
1
dt +o
1
d\
1
(t)
dA
2
(t) = j
2
dt +o
2
d\
2
(t)
Brownian Motion represents a source of risk (factor) and n is called factor loading.
(a) Construct a Stochastic Dierential Equation for 1 (t). Which value of n would make 1 (t) a
martingale?
Since A
1
(t) and A
2
(t) are independent, we can use an It-linear combination of dA
1
and dA
2
to construct an SDE for a two-dimensional Brownian Motion (\
1
t
. \
2
t
)
dA
1
(t) = j
1
dt +o
1
d\
1
t
dA
2
(t) = j
2
dt +o
2
d\
2
t
=
d1 (t) = n
_
j
1
dt +o
1
d\
1
t
_
+
_
1 n
2
_
j
2
dt +o
2
d\
2
t
_
= to simplify lets dene ~ o
1
= no
1
and ~ o
2
=
_
1 n
2
o
2
= (nj
1
+
_
1 n
2
j
2
)dt +
n=2

i=1
~ o
i
d\
i
t
The expression was optimised to show that we are only concerned with the analysis of the drift
term, regardless of whether BM is one, two or n-factor. 1 (t) can only be a martingale if
n =
j
2
_
j
2
1
+j
2
2
(b) Evaluate the variance V[1 (t)] and provide the distribution for the increment of 1 (t). Does 1 (t)
keep the properties of the Brownian Motion?
Consider the variance of 1 (t) = nA
1
(t) +
_
1 n
2
A
2
(t)
V[1 (t)] = V[nA
1
(t) +
_
1 n
2
A
2
(t)]
= n
2
V[A
1
(t)] + (1 n
2
) V[A
2
(t)] moving constant out of variance operator
= n
2
t + (1 n
2
)t = t variance of standard BM scales with time
It follows that the increment 1 (t) 1 (:) will be distributed as ~ `(0. t :) = `(0. t).
As a linear combination of two Brownian Motions, 1 (t) keeps the properties of the Brownian
Motion: the increment of standardised 1 (t) 1 (:). \: < t follows the distribution ~ `(0. t),
which is the sum of `(0. tj
2
) and `(0. t(1 j
2
)).
2. Assume that an asset price o evolves according to the SDE
do
o
= (j 1) dt +odA
where j and o are constants. In addition o pays out a continuous dividend stream equal to 1odt
during the innitesimal time interval dt, where 1 the dividend yield is constant.
Now suppose a European style derivative security \ (o. t) is written on this asset with the properties
that at expiry the holder receives the asset and prior to expiry the derivative pays a continuous cash
ow C (o. t) dt during each time interval of length dt.
2
(a) Show that the option price satises
J\
Jt
+
1
2
o
2
o
2
J
2
\
Jo
2
+ (: 1) o
J\
Jo
:\ = C (o. t) .
(You are required to derive this PDE)
= \ o
d = d\ do 1odt +C (o. t) dt.
because we are short the stock. Using the earlier hedging argument gives
d =
_
J\
Jt
+jo
J\
Jo
+
1
2
o
2
o
2
J
2
\
Jo
2
_
dt +oo
J\
Jo
dA
(jodt +oodA) 1odt +C (o. t) dt
Put =
@V
@S
to eliminate risk and use no-arbitrage
d =
_
J\
Jt
+
1
2
o
2
o
2
J
2
\
Jo
2
1o
J\
Jo
+C (o. t)
_
dt
= :
_
\ o
J\
Jo
_
dt
from which the BSE is obtained
J\
Jt
+
1
2
o
2
o
2
J
2
\
Jo
2
+ (: 1) o
J\
Jo
:\ = C (o. t)
(b) Suppose that the cash ow C (o. t) in part a. has the form C (o. t) = , (t) o. By writing
\ = (t) o and assuming a nal condition at time 1 given by
\ (o. 1) = o.
show that the delta of the derivative security is
(o. t) = c
D(Tt)
+
_
T
t
c
D(t)
, (t) dt.
(You are required to solve the PDE together with the nal condition)
Writing C (o. t) = , (t) o gives
J\
Jt
+
1
2
o
2
o
2
J
2
\
Jo
2
+ (: 1) o
J\
Jo
:\ = , (t) o
and we now use the transformation \ = c(t) o to convert to an ode which is a function of t
alone.
J\
Jt
=
0
(t) o ;
J\
Jo
= ;
J
2
\
Jo
2
= 0
For the nal condition we know
\ (o. 1) = o = (1) o
== (1) = 1
3
So the original problem reduces to
d
dt
+ (: 1) : = , (t)

d
dt
1 = ,
which is a rst order linear equation (i.e. integrating factor method). I.F is
c
Dt
so the ode becomes
c
Dt
d
dt
1c
Dt
= ,c
Dt
d
dt
_
c
Dt

_
= ,c
Dt
_
T
t
d
_
c
D
(t)
_
=
_
T
t
, (t) c
D
dt
_
c
D
(t)
_

T
t
=
_
T
t
, (t) c
D
dt
c
DT
(1) c
Dt
(t) =
_
T
t
, (t) c
D
dt
and we know (1) = 1. hence
c
DT
c
Dt
(t) =
_
T
t
, (t) c
D
dt
c
Dt
(t) = c
DT
+
_
T
t
, (t) c
D
dt
(t) = c
D(Tt)
+
_
T
t
, (t) c
D(t)
dt
So the option price \ (o. t) = (t) o and (o. t) =
J\
Jo
= (t) =
c
D(Tt)
+
_
T
t
, (t) c
D(t)
dt
3. Consider the function , () = c
y
and 1 ~ `(0. 1). Show that the variance of , () is
o
2
f
= c (c 1) .
o
2
f
= 1[ ,
2
(r)] 1
2
[ , (r)]
_
2:1[ , (r)] =
_
1
1
c
x
c
x
2
=2
.dr =
_
1
1
exp
_

1
2
_
r
2
+ 2r
__
.dr
=
_
1
1
exp
_

1
2
_
(r + 1)
2
1
__
.dr = c
1=2
_
1
1
exp
_

1
2
(r + 1)
2
_
.dr
= c
1=2
_
1
1
exp
_

1
2
n
2
_
.dn where n = r + 1 dn = dr
4
Therefore
_
2:1[ , (r)] = c
1=2
_
2:
1[ , (r)] = c
1=2
.
A similar type of working gives
_
2:1[ ,
2
(r)] =
_
1
1
c
2x
c
x
2
=2
.dr =
_
1
1
exp
_

1
2
_
r
2
+ 4r
__
.dr
= c
2
_
1
1
exp
_

1
2
(r + 2)
2
_
.dr
= c
2
.
Then
o
2
f
= c
2
c = c (c 1) .
4. The "Speed" of an option C (o. t) is given by
Speed =
J
3
C
Jo
3
.
If o = :oo and t = :ot , by obtaining 3 suitable Taylor expansions for the option price C
m
n
show
that a Finite Dierence Approximation for the Speed is given by
J
3
C
Jo
3
-
1
oo
3
_
c
1
C
m
n+2
+c
2
C
m
n+1
+c
3
C
m
n
+c
4
C
m
n1
_
.
where the values of the constants c
i
for i = 1. 2. 3. 4 must be obtained.
C (o +oo. t) = C +C
S
oo +
1
2
C
SS
oo
2
+
1
6
C
SSS
oo
3
+C
_
oo
4
_
(1)
C (o oo. t) = C C
S
oo +
1
2
C
SS
oo
2

1
6
C
SSS
oo
3
+C
_
oo
4
_
(2)
C (o + 2oo. t) = C + 2C
S
oo + 2C
SS
oo
2
+
4
3
C
SSS
oo
3
+C
_
oo
4
_
(3)
(1) (2) gives:
C (o +oo. t) C (o oo. t) = 2C
S
oo +
1
3
C
SSS
oo
3
(A)
(3) (A) gives:
C (o + 2oo. t) C (o +oo. t) +C (o oo. t) = C (o. t) + 2C (o +oo. t) 4C (o. t) + 2C (o oo. t)
+C
SSS
oo
3
In the step above we have also used C
SS
oo
2
= (C (o +oo. t) 2C (o. t) +C (o oo. t))
Rearranging gives
C
SSS
=
1
oo
3
C (o + 2oo. t) 3C (o +oo. t) + 3C (o. t) C (o oo. t)
which is an approximation to a third order partial derivative.
Recall: if C = C (o. t), then in nite-dierence form, this can be written as C
m
n
. Hence C (o +oo. t) =
C
m
n+1
; C (o oo. t) = C
m
n1
; C (o + 2oo. t) = C
m
n+2
and we can write
5
Speed -
1
oo
3
_
C
m
n+2
3C
m
n+1
+ 3C
m
n
C
m
n1

.
5. The BlackScholes formula for the value of a put option 1 (o. t) is
1 (o. t) = 1 exp(:(1 t))`(d
2
) o`(d
1
)
From this expression, nd the BlackScholes value of the put option in the following limits:
(a) (time tends to expiry) t 1 , o 0;
o
1
1:
d
1
d
2
_
=
` (d
1
)
` (d
2
)
_
0 therefore 1 0
o
1
= 1:
d
1
d
2
_
0 =
` (d
1
)
` (d
2
)
_

1
2
therefore 1 0
o
1
< 1:
d
1
d
2
_
=
` (d
1
)
` (d
2
)
_
1 therefore 1 1 o
(b) (volatility tends to zero) o 0, t < 1;
log
_
o
1
_
+: (1 t) 0:
d
1
d
2
_
=
` (d
1
)
` (d
2
)
_
0 therefore 1 0
log
_
o
1
_
+: (1 t) = 0:
d
1
d
2
_
0 =
` (d
1
)
` (d
2
)
_

1
2
therefore 1 0
log
_
o
1
_
+: (1 t) < 0:
d
1
d
2
_
=
` (d
1
)
` (d
2
)
_
1 therefore 1 1c
r(Tt)
o
(c) (volatility tends to innity) o , t < 1
d
1
=` (d
1
) 0
d
2
=` (d
2
) 1
_
therefore 1 1c
r(Tt)
Section B
6. Find the implied volatility of the following European call option. The call has four months until
expiry and an exercise price of $100. This call is worth $6.51 and the underlying trades at $101.5,
discount using a short- term risk-free continuously compounding interest rate of 8% per annum. You
can use Excel or Matlab.
6.51 = 101.5 ` (d
1
) 100 c
0:080:33
` (d
2
) . (c)
where d
1
=
log (101.5,100) +
_
0.08 +
1
2
o
2
_
0.33
o
_
0.33
and d
2
= d
1
o
_
0.33.
This is a root nding problem. Use an iterative scheme to solve (c) for the unknown volatility o
imp
(e.g. Solver in Excel). The implied vol o
imp
= 17.99%.
6
7. Consider evaluating the Gaussian integral
J =
1
_
2:
_
1
1
r
2
exp
_
r
2
,2
_
dr
by writing this as
J = E[A
2
] ~ J
N
=
1
`
N

n=1
r
2
n
.
r
n
follows a standard normal distribution, i.e. r
n
~ `(0. 1).
Implement this method in VBA, C++ or Matlab and examine J
N
for dierent values of `.
Plot an error graph (dierence between exact and numerical values).
Any graph like this that shows the error decreasing as the number of random variables increases.
Also good if some mention of why the exact value is one.
7
8. Describe in detail a Monte Carlo simulation algorithm for estimating
o =
_
1
0
c
x
2
dr.
Hint: You are required to rst obtain a suitable transformation to convert this to
an integral problem over (0. 1) . Implement your algorithm in VBA, C++ or matlab
for dierent numbers of random numbers generated and comment on how accuracy is
achieved.
Let n = 1, (1 +r) . then we have r
2
= (1 n)
2
,n
2

du
dx
= 1, (1 +r)
2
= n
2
. Hence the integral
becomes
o =
_
0
1

exp
_

_
1
u
1
_
2
_
n
2
dn =
_
1
0
exp
_

_
1
u
1
_
2
_
n
2
dn
= E
_
c
(1U)
2
=U
2
l
2
_
where l ~ l (0. 1) . So the algorithm becomes
Generate l
1
. l
2
. ....... l
n
~ l (0. 1)
Calculate
A
i
=
1
l
2
i
c
(1U
i
)
2
=U
2
i
Finally set

o =
1
:
n

i=1
A
i
There is a closed form solution to the integral which is
_
:,2 so a graph can be presented of how the
numerical value varies with the number of simulations used.
Mark Scheme:
Question 1 2 3 4 5 6 7 8
Marks 15 15 10 10 10 10 10 20
8

Вам также может понравиться