11.

7 Problems

i.i.d. RVs with EX1

=
1, Var (X1) =
a2< oo. Put So:=0,
X2,.. be
1. Let X1,
Xn, n 1 . Compute
Sn= Sn-1 +
m, n = 0, 1,2,..
(i) E (S,+ m Sn),
(ii) E (X1|S,),
n > 1;
n, n
=
0, 1,2,.
(ii) E (S+m|S,),
n..
(iv) E (Sm S,), m
=
0, 1,.. .
,

Hints. (ii) E (X1| S,) =

E(AX2| Sn) (n2 2) etC. by symmetry. (iv) Yo may
wish to use the result of one of the parts (i)-(iii) above.
rateA>0, F =o{N,, 0<s < the with
2. Let {N,}+>0 be a Poisson process
the time t. Using the properties of the Poiseon
history" of the process up to
process and conditional expectations, find

(i) E (N+|F), s, t 2 0;

(i) E (Ns| F), s, t 20

(ii) E (N.|F) and E (N|Fi), 0 Ss<t
(iv) E (N|N:) and E (N;|N.), 0<sSt.
Hint. It is not much different from the previous problem, is it?
3. Let {Xi}t=0,1,.,T be a positive SP adapted to a filtration F = {Fi. ln eactn
of the following cases, say if the RV T is an ST w.r.t. F (if the condition m the
definition of the random time r in (ii)-(iv) is never met for t S T, we
just put
put
T:=T t0 avoid any inconvenience). Explain (e.g., expressing ts t"
in terms of the RVs Xk, k =
1,2,...).
(i)T=m = const;
(ii) T = Ti
AT2, where T are STs, j =
1,2;
(ii) T= min{t 2 0: X1+1/Xt>
1};
(iv) T= min {t2 0: 2k-
Xk Xi};
lrmaxttST: X, > 10
L eI
r } integrable RV (i.e., EY| < o) on
be an integrable

F={Fi}:20.P) Show that X filtered probability a

F
(Q2 :=E(Y|F) is an MG.6 space
1,0be square-integrable (i.e., EXf <
et {Xfe>o a

orthogonal increments in the sense that,o)for MG. Show that the

has orthogona
proces has
ta.one has E (X- X,)(X, Xt,) =0.
one any 0 ti ta-

So 0. Sn =
1et
Le
lar ( ) = o 2<
Iif**** In, n 2
RVs with EY, 0 1, Y; being i.i.d.
Show that Xn := S
00. =

neet the filtration Fn

-no", n>0, is an MG (i) with
respect to o(Y1,..., Yn); (ii) with respect to the natural
=

fltration no(A1,.,Xn).
Hint. Use the result of (i) and the fact that F,
(ii) C Fn (why does the last
relation hold?).
T. Denote by {N}i2o a Poisson process with rate A >0. Show that aill three
processes i) N-At; (i) ( - M) - t ; (ii) exp{uN - t(e" - 1)} (u is a
fxed real number) are MGs w.r.t. the filtration Fn = o(Ns, s < t).
S. Let So = 0, Sn = Yi t + Yn, n 2 1, Y; being ii.d. RVs with P (Y1 =
1) = 1 -P (Yi = -1) = 1/2. Denote by T:= min{n 20: Sn = a or Sn= 6}
the first time the RW S,, hits one of the (integer) barTiers a < 0 < b. Use
Theorem 11.2, without verifying its conditions in detail, to:

i) find the distribution of ST;"

i) compute ET.
and the MG
Hints. (i) Use the martingale Xn := Sn. (ii) Use the result of (i)
from Problem 6.
each play (in which you
.Suppose you are game" betting $1 at
playing a "fair
fortune after
independently of the past). Then {Xn : your
win/lose w.p.,
MG
F F1,F2,.. .j.
w.r.t. its "history"
=
{JFo,
nplaysn>0 is an
stake can be a n arbitrary
each play n =1,2,..., your
NOW suppose that, for have to decide how much
to stake before that play,
boUnded amount Yn, but you Mathematically, this
1 (inclusive).
basing on the history up to play n -
and {Ynfn>1 is a
predictable
Const < 0 ,
|Ynl Cn =

eans that, for any n,

process (cf. p. 307). on play n (as it
was Xn-Xn-1
n that case, you win the
amount Yn (Xn-Xn-1) total net gain after n plays is
and hence your
When staking $1 each time),

n= 1,2,...; Zo =0.
Zn = Y(X -X*-1),
k=l show that Xt > E(YIF)
One can
martingale.
1S NOT the as
is referred to a s a Lévy
same
IG Speaking, o
(generally
:=o(U. >Ft) MG for s o m e Y.
Sas t 0 , where F MG is the Lévy
In fact, any
uniformly integrable problem using a different approach
n0t'). (gambler's ruin)
17N nat we already solved that
e case in (3.44))
in Example 3.21 (see the second
 is called the martingale transforn!8 of
The process {.\,,}no X by Y
MG w.r.t. F.
Show that {Z»}n>o is an
(i)
The betting strategy
ot donbling when losing is called" called "nmartingale", Yom
(i)
begin with a
unit stake, win cach play w.p. (regardless of the results
and double your stake for the next
next nlas of
all the previous plays), when play
After each win. you reset the stake size to one. your net Represent gain losing,
ing
2when using this strategy as a martingale transform: specifv tho.
eSSes
and {»
(ii) Under the assumptions of part (i), assume that you stop at tho t:
the first win. Find the distribution and expectation of the ST T and v .
time
the statenment of Theorem 11.2 holds for the process {Z,} and ST T. verify if
(iv) Compute the expectation E(Z; T > n) and hence verify if the gen
eneral
condition (11.11) sutficient for (11.9) is satisfied in this case.

10. Let {S.}n2o be RW defiued follows: starting at S0 =0, the

an as

particle at each transition goes up 1 w.p. p =$ and down 2 w.p. 1-p= ng

i) Show that Xn :=27°", n =0, 1,2,..., is an MG.

ii) Introduce the STr= min{n 2 0: Xn <0.1. Use Theorem 11.2

(without
verifving its conditions) and the MG {Zn = Sn - E S,} to compute Er.

Hint. (i) First express the ST T in terms of the random walk

Sn. What are
the possible values
of S,?
Note that if Sn > Sn-1, then Sn = Sn-1 +1.
11. Let {Sn }n20 be a simple RW:
starting at some initial point So. the walking
particle at each transition goes up 1 w.p. p E
Assume that (0, 1) and down l w.p. =l-p
p # .
i) Show that {X, (a/p)°" } is an MG.
:=

(11) Suppose the

walk starts at So 0 and =
stops at the time T= min/n2
Sna Or Sn 6}, where a <0 <b are integers. Use Theorem
=

11.2 he
MG from part (i) to find the ana 11.2
and the MG distribution of RV the ST, and then 'Theore
{Z, = Sn -

n(p -

q)} to E
12. Show that the set compute T.

dense in [0,1 with {U1, U2,...}, where U; are i.i.d. U(0, 1)-RVs, is every where
probability 1.
Hnt. You may wish
to use the
13. Find the Glivenko-Cantelli theorem (2.8/).
14. Find the
distribution of X :=
2W, - Wig, 0< ti < ta.
distribution of X:= Wo +
15. Derive the W2- W3 + 2W4.
BM's FDD
and the
observation thatdensity f ri,... ,#n) see (11.22)) using (2.33)
n) (see
(Wt,
transformation of the vector Wt,.. ,W,) is the result
o simple) line
linear
result of simple a
(W Wta, Wt wiu
- with

SMartingale transforms are discrete

portant role in the
Wt., Wtn -

analogues of stochastic
mathematical theory play
an im

of finance
in
integrals ana
P
discrete time.
 independent mDonents, so that the latter vector's
density is just the product
idepenuies of the increments W- Wi-,
of the da
k =
1,2,... , n.
the transformation of the standard normal vector ZE R" into the
1
Note
No

vector of
the values of the standard BM from the simulation algorithm on
ritten
in the matrix form as (W..., W,) = ZA, A E R"xn
be writ
320 c a n
Specify the matrix A. How will the above matrix representation change if
were to directly simulate the vector (Xt1,...,Xt,), where {X} is the
you
arithmetic BM (11.16)?

17 Compute the joint

densities of (i) (2W3, W5) and (ii) (W2,2W3, Ws).

Ise Theorem 11.4 to show that {W= tW1/t}tzo is a standard BM process

that {Wijt20 is such.

proved
min{t > 0: tya+ bt } be the first time the BM crosses one
Wi =
19. Let T :=
where a > 0 and be (0,1)
ofthe two parabolic boundaries tya+ bt, t 2 0,
ET.
are some
constants. Use Theorems 11.6 and 11.2 to compute
BM crosses

20. Denote by T = min{t Wi< 2t-4} the first time the

> 0:
process
Brownian
three martingales of the
theboundary ve = 2t-4, t 20. Using the
the conditions of the
motion (Theorem 11.6) and Theorem 11.2 (do not verify
time r its8:
theorem), compute for the stopping
i) mean value ET;

i) variance Var (T);

Ee**", s 0 .
(11) Laplace transform l(s) =
also E W, and E WG.
(iv) Compute time thestandard BM
Wi = a or Wi =b} the first
min{t > 0: the three martingales
Denote by T (a < 0< 6). Using
Z1. :=
a or b
process takes one of the values not verify the conditions of
the theorem),
and Theorem 11.2 (do
of the BM
distribution of W7;
i) find the
value ET;
compute the
mean

(ii) = Ee*', s 0 , in the case when

transtorm lr(s)
Laplace
(iii) compute the
a = -1, b = 1; 1.
the mean E r when a =-1, b =

to compute

(iv) use the result of part (iii) for question (ii).

with that
the result
Compare -1,
Var (7) when
=
a
the variance
to coinpute
result of part (11)
(v) Use the
b = 1. common filtered prob-
on given on a
j0, T|
and g1 be simple proceSses direct calculatiou
22. Let ft {Wi}i20 given on
it. Show by a

ability space,
with a BM
=
Jo E Sg, ds, t e [0,T].
that E I(f)L{9)
23. Let g be a non-random
finction on [0,7| satisfying ad
fdt . Use Itós
formula to show that the following process is an MG

Y expg,dW,-9 d s , 0
te 0,T].

24. Compute the stochastic differential d cos(W:).

25. Put X, :=t+ W, t 0.

(i) Apply Itô's formula to compute the stochastic differential de-2x,

i) Is the process Y:=ea"t, t 2 0, a martingale? Explain.
26. The price St of a risky asset evolves according to the SDE

dS =
0.2S,dt +StdW, t20, 5
(this is special case of the so-called Black-Scholes framnework
a

in Chapter 13).
to be discised

(i)It is suspected that the SDE has a solution

of the form St =
where a, b and c are some constants. Use ceat+6W
ltö's formula to verify this
and find the values of the constants suspicion
a, b and c in the solution.
ii) Show that the process Xt =
1/S; satisfies the SDE

dX 0.8X;dt - XtdWi, t>0,

=

Xo =
0.2.
27. The "stochastic volatility" Heston model
{Ve}r20 follows the SDE assumes that the "variance procesS

dV =
(1 V)dt
-

+2vV dWi, t20, Vo =1.

i) Derive an SDE for
initial condition for thethe "volatility process" Z; =
V t>0, and hna
SDE (i.e., the value
Zo).
(1) Show that the SP
initial condition Z=et/2(1+ Ce/2 aw,) satisfies the SDE and
you derived in
part (i).