Martingales
Exercises
Exercise 4.1. Show that a sum of martingales is a martingale.
Exercise 4.2.
a) Is any Markovian process is a martingale? If yes, prove it. Otherwise, construct a counter-
example.
b) Is any martingale is Markovian? If yes, prove it. Otherwise, construct a counter-example.
Exercise 4.3. Let M = fMt : t 2 f0; 1; 2; :::gg be a square-integrable martingale1 existing on
a …ltered probability space ( ; F; F; P), where F is the …ltration …ltration fFt : t 2 f0; 1; 2; :::gg.
The predictable process = f t : t 2 f0; 1; 2; :::gg is constructed on the same space, that is,
for all t 2 f1; 2; :::g, t is Ft 1 measurable, 0 being F0 measurable. Moreover, we assume
that for all t 2 f0; 1; 2; :::g, the random variable t is square-integrable,that is, E [j 2t j] < 1:
Show that the process N = fNt : t 2 f0; 1; 2; :::gg de…ned as
X
t
Nt = N 0 + k (Mk Mk 1 )
k=1
is a martingale provided that N0 is F0 measurable.
Exercise 4.4. Let X be a square-integrable random variable EP [jXj] < 1 constructed on
the …ltered probability space ( ; F; F; P). Prove that the stochastic process fMt : t 2 f0; 1; 2; :::gg
de…ned as
Mt = EP [X jFt ] , t 0
is a martingale.
Exercise 4.5. Consider the probability space ( ; F; P) on which two …ltrations are con-
structed, fFt : t 0g and fGt : t 0g ; satisfying
Ft Gt
a) Let M = fMt : t 0g be a fFt : t 0g martingale and let N = fNt : t 0g be a
fGt : t 0g martingale. Is M a fGt : t 0g martingale ? Is N a fFt : t 0g martingale ?
Justify.
b) Let be a fFt : t 0g stopping time and be a fGt : t 0g stopping time. Is a
fFt : t 0g stopping time ? Is a fGt : t 0g stopping time ? Justify.
1
8t 0; E Mt2 < 1
1
Exercise 4.6. Let "1 ; "2 ; ::: be a sequence of independent random variables with zero mean
and variance Var ["i ] = 2i . Let
X
n X
n
Sn = "i and Tn2 = 2
i:
i=1 i=1
Prove that fSn2 Tn2 gn2N is a martingale.
Exercise 4.7. Let fXt : t 0g be a fGt : t 0g martingale and Ft = (Xs ; s t). Prove
that fXt : t 0g is also a fFt : t 0g martingale.
2
Solutions
1 Exercise 4.1
Let X = fXt : t 2 f0; 1; :::gg and Y = fYt : t 2 f0; 1; :::gg, two martingales on ( ; F; F; P) :
Since 8t 2 f0; 1; :::g,
EP [jXt + Yt j] EP [jXt j] + EP [jY j] < 1;
| {z } | {z t }
<1 <1
since X is a martingale since Y is a martingale
the (M 1) condition is veri…ed.
Since two Ft measurable random variables is Ft measurable, then 8t 2 f0; 1; :::g, Xt +Yt
is Ft measurable. Therefore, the stochastic process X + Y is adapted to the …ltration F:
Condition (M 3) : 8s; t 2 f0; 1; :::g such that s < t
EP [Xt + Yt jFs ] = EP [Xt jFs ] + EP [Yt jFs ] = Xs + Ys :
The proof can be generalized to the sum of more than 2 martingales using induction.
2 Exercise 4.2
a) It is not all Markovian processes that are martingales.
Counter-example. On the probability space ( ; F; P),consider the random walk X
de…ne as
X
t
X0 = 0 and 8t 2 f0; 1; 2; :::g , Xt = n
n=1
where the sequence of independent and identically distributed random variables f t : t 2 f1; 2; :::gg
has a positive expectation EP [ t ] = > 0. Since X is a random walk, it is Markovian. But
X cannot be a martingale since its expectation varies over time. Indeed,
" t #
X Xt X
t
EP [Xt ] = EP n = E P
[ n ] = =t :
n=1 n=1 n=1
3
b) It is not all martingales that are Markovian. Counter-example :
! X1 (!) X2 (!) X3 (!) P
1
!1 1 2 4 8
1
!2 1 2 0 8
1
!3 1 0 0 8
1
!4 1 0 0 8
1
!5 1 0 2 8
1
!6 1 0 2 8
1
!7 1 2 2 8
1
!8 1 2 2 8
Indeed,
fX1 g = ff! 1 ; ! 2 ; ! 3 ; ! 4 g ; f! 5 ; ! 6 ; ! 7 ; ! 8 gg
fX1 ; X2 g = ff! 1 ; ! 2 g ; f! 3 ; ! 4 g ; f! 5 ; ! 6 g ; f! 7 ; ! 8 gg
fX1 ; X2 ; X3 g = ff! 1 g ; f! 2 g ; f! 3 ; ! 4 g ; f! 5 g ; f! 6 g ; f! 7 ; ! 8 gg
The process X = fXt : t 2 f1; 2; 3gg is not Markovian on the space ( ; F; P) where F is the
algebra generated by all elements of : Indeed,
P (X3 = 2 and X2 = 0)
P (X3 = 2 jX2 = 0) =
P (X2 = 0)
1
P f! 5 g 8 1
= = 1 =
P f! 3 ; ! 4 ; ! 5 ; ! 6 g 2
4
but
P (X3 = 2, X2 = 0 and X1 = 1)
P (X3 = 2 jX2 = 0 and X1 = 1) =
P (X2 = 0 and X1 = 1)
P (?) 0
= = 1 = 0:
P f! 3 ; ! 4 g 4
However, the process X = fXt : t 2 f1; 2; 3gg is a martingale on the space ( ; F; P) if the
…ltration is generated by the process itself. Indeed, the condition (M 1) is trivially satis…ed,
4
(M 2) is also satis…ed from the construction of the …ltration. Need to verify the last condition
(M 3 ). But
EP [X3 j fX1 ; X2 g] = X2
since
1 1
4 +0
8! 2 f! 1 ; ! 2 g , EP [X3 j fX1 ; X2 g ] (!) = 8
1
8
= 2 = X2 (!)
4
1 1
P 0 8
+0 8
8! 2 f! 3 ; ! 4 g , E [X3 j fX1 ; X2 g] (!) = 1 = 0 = X2 (!)
4
1 1
P 2 8
2 8
8! 2 f! 5 ; ! 6 g , E [X3 j fX1 ; X2 g] (!) = 1 = 0 = X2 (!)
4
1 1
P 2 8
+2 8
8! 2 f! 7 ; ! 8 g , E [X3 j fX1 ; X2 g ] (!) = 1 = 2 = X2 (!)
4
and
EP [X2 j fX1 g] = X1
since
8! 2 f! 1 ; ! 2 ; ! 3 ; ! 4 g
P 2 81 2 81 + 0 1
8
+0 1
8
E [X2 j fX1 g ] (!) = 1 = 1 = X1 (!)
2
8! 2 f! 5 ; ! 6 ; ! 7 ; ! 8 g
0 81 + 0 81 + 2 1
+2 1
EP [X2 j fX1 g ] (!) = 1
8 8
= 1 = X1 (!) :
2
5
3 Exercise 4.3
Veri…cation of (M 1) :
" #
X
t
E [jNt j] = E N0 + k (Mk Mk 1 )
k=1
X
t
E [jN0 j] + E [j k j jMk Mk 1 j]
k=1
X
t
1=2 1=2
E [jN0 j] + E j k j2 E jMk Mk 1 j2
k=1
from Cauchy-Schwarz inequality
X
t
1=2 1=2
= E [jN0 j] + E j k j2 E Mk2 2Mk Mk 1 + Mk2 1
k=1
X
t
1=2
= E [jN0 j] + E j k j2 E Mk2 2E [Mk Mk 1 ] + E Mk2 1
k=1
X
t
1=2
= E [jN0 j] + E j k j2 E Mk2 2E [Mk Mk 1 ] + E Mk2 1
k=1
since E [Mk Mk 1 ] = E [E [Mk Mk 1 jFk 1 ]]
= E [Mk 1 E [Mk jFk 1 ]]
= E Mk2 1
X
t
1=2 1=2
= E [jN0 j] + E j k j2 E Mk2 E M2
| {z }
k=1
| {z }| {z k 1
}
<1 <1 <1
< 1
Veri…cation de (M 2) :
X
t
Nt = N0 + k (Mk Mk 1 ) est Ft mesurable.
|{z} | {z }
k=1 F mesurable Fk mesurable
k 1
| {z }
Fk mesurable donc Ft mesurable
6
Veri…cation de (M 3) : …rstly, note that for all 0 s < t < 1;
X
t
Nt = Ns + k (Mk Mk 1 ) :
k=s+1
Indeed,
X
t
N t = N0 + k (Mk Mk 1 )
k=1
Xs X
t
= N0 + k (Mk Mk 1 ) + k (Mk Mk 1 )
k=1 k=s+1
X
t
= Ns + k (Mk Mk 1 ) :
k=s+1
Therefore,
" #
X
t
E [Nt j Fs ] = E Ns + k (Mk Mk 1 ) Fs
k=s+1
X
t
= Ns + E[ k (Mk Mk 1 )j Fs ]
k=s+1
X
t
= Ns + E [E [ k (Mk Mk 1 )j Fk 1 ]j Fs ] from (EC3)
k=s+1
2 3
X
t
= Ns + E4 k E [(Mk Mk 1 )j Fk 1 ] Fs 5
| {z }
k=s+1 =0
= Ns :
7
4 Exercise 4.4
Veri…cation of (M 1) :
EP [jMt j] = EP EP [X jFt ]
EP EP [jXj jFt ] since X jXj
= EP EP [jXj jFt ] since EP [jXj jFt ] 0
= EP [jXj] since (EC3)
< 1 by hypothesis
Veri…cation of (M 2) : Mt = EP [X jFt ] is Ft measurable because constant on the atoms
Ft .
Veri…cation of (M 3) : Let 0 s t.
EP [Mt jFs ] = EP EP [X jFt ] jFs from the defn. of Mt
= EP [X jFs ] from (EC3) since Fs Ft
= Ms from the de…nition of Ms .
