Stochastic Calculus
Example sheet 3 - Lent 2015
Michael Tehranchi
Problem 1. Let X be a continuous local martingale with X0 = 0. For each s 0, let
T (s) = inf{t 0 : hXit > s}.
Show that the stopped process X T (s) is a square-integrable martingale. Hence, conclude that
for almost every {hXi < } the limit limt Xt () exists.
Problem 2. Let P Q. Show that Xn X in P-probability if and only if Xn X in
Q-probability.
Problem 3. (Yet another martingale representation) Let X be a real continuous local martingale whose quadratic variation is
Z t
hXit =
as ds
0
for a non-negative, predictable process (at )t0 . Show that there exists a real Brownian motion
W (possibly defined on an extended probability space) such that
Z t
Xt =
as dWs
0
How could this be generalised to higher dimensions?
Problem 4. Let W be a standard two-dimensional Brownian motion, and let a R2 ,
kak = 1 be a constant. Let Zt = log kWt ak and set T0 = 0 and
Tk = inf{t Tk1 : |Zt ZTk1 | > 1}
(a) Show that for every k, the process (ZtTk ) is a martingale.
(b) Let Xk = ZTk . Use the optional stopping theorem to show that (Xk )k0 is a simple
symmetric random walk on the integers.
(c) Conclude that for a two-dimensional Brownian motion W , the set {t 0 : kWt ak < r}
is unbounded a.s. for any a R2 and r > 0.
Problem 5. Let W be a three-dimensional Brownian motion, and let N3 (0, I) be
independent of W . Using the fact that P(Wt = for any t 0) = 0, we define a positive
process X by
1
Xt =
kWt k
(a) Use Itos formula to show that X is a local martingale. Indeed, show that there exists a
real Brownian motion B such that
dXt = Xt2 dBt .
(b) Compute E(Xt ). Why does this show that X is a strictly local martingale?
(c) Compute E(Xt2 ). Conclude that X is a uniformly integrable local martingale. Compare
this to example sheet 2 problem 1.
(d) Verify that E(hXit ) = for all t > 0.
(e) Show that Xt 0 a.s. and conclude that for a three-dimensional Brownian motion W ,
we have kWt k a.s.
1
Problem 6. If M is a scalar continuous local martingale with hM it a.s., show that
Mt /hM it 0. Conclude that E(M )t 0 a.s.
Problem 7. Let W = (Wt )0tT be a real Brownian motion generating the filtration F,
where the non-random time-horizon T is finite. Also let be a bounded predictable process.
Show that for any FT -measurable bounded random variable there exists a constant x and
a predictable process (t )0tT such that
Z T
=x+
t (dWt + t dt).
0
Problem 8. Let W be a real Brownian motion and T a stopping time.
(a) Show that if E(T ) < then E(WT2 T ) = 0.
2
2
(b) Show that if E(e T /2 ) < for some R, then E(eWT T /2 ) = 1.
Problem 9. (OrnsteinUhlenbeck process) Let W be a real Brownian motion, and let
Z t
t
Xt = e +
e(ts) dWs
0
for some R.
(a) Verify that X satisfies the following stochastic differential equation:
dXt = Xt dt + dWt ,
X0 = .
(b) Show that
1
t
2t
Xt N e , (1 e ) .
2
2
(c) Show that exp(Xt t) is a martingale.
(d) Now suppose N (0, 2 ), independent of W . Find the unique 2 such that Xt
N (0, 2 ) for all t 0. Compute Cov(Xs , Xt ) in this case.
(e) Let 2 and be as in part (d), and let Yt = et We2t . Show that Y has the same law as
X.
Problem 10. (Stratonovich integral) Let X and Y be continuous semimartingales. The
Stratonovich integral of X with respect to Y is defined by
Z t
Z t
1
Xs dYs + hX, Y it
Xs dYs =
2
0
0
where the integral on the right side of the equation is an Ito integral. Show that if f is three
times continuously differentiable, then Itos formula can be written
df (Xt ) = f 0 (Xt ) dXt .