You are on page 1of 6

Semester 1, 2011 MTH3251/ETC3510/ETC5351

MTH3251/ETC3510/ETC5351
Exam
Semester 1, 2011
Solutions
Page 2 of 7
Semester 1, 2011 MTH3251/ETC3510/ETC5351
Throughout the paper, B
t
, t 0, denotes the Brownian motion started at 0.
1. In a game of roulette, there are 37 squares, 18 of which are red and 18 black, 1
square is neither. Suppose you bet on a colour with $1 each bet. Denote by Y
n
the outcome of the n-th bet with distribution: Y
n
= 1 with probability 18/37
and Y
n
= 1 with probability 19/37. Let X
n
be the fortune after n bets. You
start with $10, X
0
= 10, and play until no money is left or you reach $20.
(a) Show that the process M
n
= (19/18)
X
n
, n = 0, 1, . . ., is a martingale.
Integrability: |X
n
| X
0
+n. Hence M
n
(19/18)
X
0
+n
and EM
n
< .
Martingale property: as X
n+1
= X
n
+ Y
n+1
, M
n+1
= M
n
(q/p)
Y
n+1
.
Hence E(M
n+1
|M
n
) = M
n
E((q/p)
Y
n+1
) by using independence. Now,
E((q/p)
Y
n+1
) = (q/p)p + (q/p)
1
q = q + p = 1. Hence E(M
n+1
|M
n
) =
M
n
. 6
(b) Show that the duration of the game is a stopping time.
{ > n} = {X
i
> 0 and X
i
< 20 for i =, 1, . . . , n}. Therefore it is
decided by observing the rst n outcomes of the process. 2
(c) Derive the probability of reaching $20 before ruin. Hint: use (without
justication) the Optional Stopping Theorem.
By Optional Stopping Theorem, E(M

) = E(M
0
). So we have
E

(19/18)
X

= E

(19/18)
X
0

= (19/18)
10
.
Denote by a the probability P(X

= 20). Then using that is nite


P(X

= 0) = 1 a. Opening the expectation


E

(19/18)
X

= a(19/18)
20
+ (1 a)(19/18)
0
= a(19/18)
20
+ 1 a.
a(19/18)
20
+ 1 a = (19/18)
10
a =
(19/18)
10
1
(19/18)
20
1
.
7
[Total 15]
2. Denote by S
n
, n = 0, 1, 2, . . ., a Random Walk S
n
= S
0
+

n
i=1
Y
i
, where S
0
>
0, is a constant, and the Y
i
s are independent normally distributed random
variables N(,
2
) with > 0.
(a) Find a constant C > 0 such that the process M
n
= e
CS
n
, n = 0, 1, 2, . . .,
is a martingale.
Page 3 of 7
Semester 1, 2011 MTH3251/ETC3510/ETC5351
Since M
n+1
= M
n
e
CY
n+1
, C solves Ee
CN(,
2
)
= 1. Using mgf of a
normal, e
C+C
2

2
/2
= 1. C =
2

2
5
(b) Let T denote the rst time the random walk S
n
takes a negative value.
Show that for any n, P(T n) e
CS
0
.
T is a stopping time. T n is a bounded stopping time. Using the
Optional Stopping Theorem, EM
Tn
= M
0
= e
CS
0
. Now EM
Tn
=
EM
T
I(T n)+EM
n
I(T > n) EM
T
I(T n) EI(T n) because
M
T
1 by denition of T as S
T
< 0 and C < 0 so that CS
T
> 0
and M
T
= e
CS
T
> 1. Since EI(T n) = P(T n), the inequality is
proved. 5
(c) Formulate the Discrete Time Risk Model in Insurance with normally
distributed claims. Give a bound on the probability of ruin in terms of
the initial funds by using the Random Walk model.
The capital at time n, U
n
satises U
n+1
= U
n
+ c X
n+1
, where c is
the total premium and X
n+1
is payout on claims. Thus U
n
satises the
Random Walk model with Y
i
= c X
i
. Since X
i
is normal N(
X
,
2
),
Y
i
is N(c
X
,
2
). Applying the bound for ruin probability in the
RW, we obtain P(T n) e
CU
0
with C =
2(c
X
)

2
. Since the bound
does not depend on n the probability of ruin P(T < ) satises same
bound. 5
[Total 15]
3. (a) Give the distribution of B
1
.
N(0,1) 1
(b) Give the joint distribution of B
1
and B
3
.
Bivariate normal with mean 0 and covariance matrix =

1 1
1 3

. 3
(c) By quoting an appropriate result, nd the conditional expectation E(B
3
|B
1
).
E(B
3
|B
1
) = B
1
as B
t
is a martingale 3
(d) State with reason whether the process X
t
= B
t
is also a Brownian
motion.
Yes. It has independent and normally distributed increments N(0, ts).
4
(e) State with reason whether the process X
t
= B
3
t
is also a Brownian motion.
No. either: increments B
3
t
B
3
s
= (B
t
B
s
)(B
2
t
+2B
t
B
s
+B
2
s
) depend
on B
s
; or Itos formula to see that B
3
t
is not a martingale. 4
[Total 15]
Page 4 of 7
Semester 1, 2011 MTH3251/ETC3510/ETC5351
4. The process X
t
is given by X
t
=

t
0
sdB
s
, t 0.
(a) Show that X
t
, t 0 is a martingale. Give the distribution of X
t
.
Since

t
0
s
2
ds = t
3
/3 < the It o integral is a martingale. Alternatively
direct verication. Since the function under the integral is non-random
X
t
has N(0,

t
0
s
2
ds = t
3
/3) distribution. 3
(b) Show that the process e
X
t

1
6
t
3
, t 0, is a martingale. Hint: use It os
formula and check the conditions for It os integral to be a martingale.
Let U
t
= e
X
t

1
6
t
3
. Using Itos formula with f(x, t) = e
x
1
6
t
3
, we obtain
dU
t
= U
t
dX
t
= U
t
tdB
t
. Hence U
t
= 1 +

t
0
U
s
sdB
s
.
For it to be a martingale, it is enough that

t
0
E(U
s
s)
2
ds < . Now
E(U
2
t
) = Ee
2X
t

1
3
t
3
= e

1
3
t
3
Ee
2X
t
. As X
t
N(0, t
3
/3), Hence Ee
2X
t
=
e
2t
3
/3
. This gives E(U
2
t
) = e
t
3
/3
. The condition

t
0
E(U
s
s)
2
ds <
holds, as required. 7
[Total 10]
5. Assume the following market model. The stock price today is $2 and in the
next period it can be $4 or $1. The simple interest rate is over this period
is 10%, r = 1.1. State what is meant by arbitrage and show that there is no
arbitrage in this model.
Arbitrage is a self-nancing portfolio with initial value 0 and a positive
value at time 1. By the First Fundamental Theorem a model does
not have arbitrage if there is an equivalent martingale measure EMM,
that makes the discounted stock price process into a martingale. In
this model the discounted stock price process is S
0
= 2 and S
1
/r =
4/1.1, 1/1.1. The martingale condition is E(S
1
/r) = S
0
. (4/1.1)p +
(1/1.1)(1 p) = 2. This gives p = 0.4. Since p is a probability, the
EMM exists hence there is no arbitrage. 5
[Total 5]
6. Assume the following market model. The stock price evolves according to
S
t
= e
0.2t+0.2B
t
, where B
t
is Brownian motion, 0 t 1. The continuously
compounding interest in a saving account over this period is 0.1.
(a) Let X
t
denote the discounted stock price process. Find a such that e
at
X
t
is a martingale and show that there is only one such a.
X
t
= S
t
e
rt
= e
0.1t+0.2B
t
. The exponential martingale is e
0.2B
t
(0.2
2
/2)t
.
Hence a = 0.12. The martingale M
t
= e
at
X
t
= e
0.2B
t
0.02t
. If a > 0.12
then e
at
X
t
= M
t
e
(10.12)t
, which is a martingale times a decreasing
function, hence not a martingale. Similarly for a < 0.12 we get a mar-
tingale times an increasing function. Hence a is unique. Or stochastic
dierential that gives a drift. 5
Page 5 of 7
Semester 1, 2011 MTH3251/ETC3510/ETC5351
(b) Give the distribution of S
1
under the equivalent martingale probability
measure. Express the price at time t = 0 of a call option on this stock
with exercise price $10 and expiration time T = 1 as an expectation of
its payo. Explain briey how to obtain the Black-Scholes formula from
that expression.
Under Q, S
1
= e
(r
1
2

2
)+B
1
and has a Lognormal distribution
LN(0.08, 0.04). The price of option e
0.1t
E
Q
(S
1
10)
+
, where Q is
the EMM, or e
0.1t
E(S
1
10)
+
, where S
1
has above Lognormal distri-
bution. Calculating this expression the BS formula is obtained. 5
[Total 10]
7. Assume that the spot rate r
t
follows Vasiceks model, given by
r
t
= 0.1 + 0.1e
t
+e
t

t
0
e
s
dB
s
.
(a) Give the distribution of r
t
and state its mean and variance.
N(0.1 + 0.1e
t
, e
2t

t
0
e
2s
ds) = N(0.1 + 0.1e
t
,
1
2
(1 e
2t
)). 3
(b) Derive the stochastic dierential equation for r
t
.
Integration by parts gives dr
t
= (0.1 r
t
)dt +dB
t
. 3
(c) Explain briey what is meant by mean reversion and give the level to
which the spot rate r
t
reverts.
The rate reverts to level a = 0.1. When r
t
> 0.1 then the drift is
negative, when r
t
< 0.1 then the drift is positive. 3
(d) Find the covariance function of the process r
t
, t 0. Hint: Show that the
process X
t
= r
t
e
t
0.1e
t
0.1 is a martingale thus obtain the covariance
function of X
t
.
X
t
= r
t
e
t
0.1e
t
0.1 =

t
0
e
s
dB
s
. Hence X
t
is a martingale as
an Ito integral. For u < t, Cov(X
t
, X
u
) = E(X
t
X
u
) EX
t
EX
u
=
E(X
t
X
u
) = EE(X
t
X
u
|X
u
) = EX
2
u
= E(

u
0
e
s
dB
s
)
2
=

u
0
e
2s
ds =
1
2
(e
2u
1). r
t
= (X
t
+ 0.1)e
t
+ 0.1. Using properties of covariance
Cov(r
t
, r
u
) = e
t
e
u
Cov(X
t
, X
u
) =
1
2
e
t
(e
u
e
u
). 6
[Total 15]
8. Explain how to do the following and justify your choice.
Page 6 of 7
Semester 1, 2011 MTH3251/ETC3510/ETC5351
(a) How to simulate an observation from a Pareto distribution with the cu-
mulative distribution function F(x) = 1 2/x, x 2, and F(x) = 0 for
x < 2.
Use the inverse function method. F(x) = 12/x = y. x = 2/(1y), so
that F
1
(x) = 2/(1x). Generate U uniform (0, 1) then X = 2/(1U)
has the required distribution. Check: P(X x) = P(2/(1U) x) =
P(U 1 2/x) = F(x). 5
(b) How to simulate an observation from a bivariate Normal distribution with
mean 0 and covariance matrix =

1 1
1 4

. In other words, simulate a


pair of correlated normal random variables with these means and covari-
ances.
Simulate two independent standard Normal Z
1
and Z
2
. Then use the
transformation X
1
= Z
1
, X
2
= Z
1
+

3Z
2
. Then the transformation
is given by multiplication by A =

1 0
1

. Clearly, AA
T
= .
Therefore X
1
, X
2
are bivariate Normal as required. 5
(c) How to evaluate the integral

1
0
dx
log(x+2)
by simulations.
I =

1
0
dx
log(x+2)
= Eh(U), where h(x) =
1
log(x+2)
and U is uniform (0, 1).
Hence I
1
N

N
i=1
h(U
i
), where U
i
are i.i.d. Uniform. Justication is
given by the Law of Large Numbers,
1
N

N
i=1
h(U
i
) Eh(U) as N .
5
[Total 15]
End of examination questions
End of examination paper
Page 7 of 7

You might also like