processes
Stochastic processes Filtration Stopping time References
Stochastic Processes
80-646-08 Calcul stochastique
Geneviève Gauthier
HEC Montréal
processes
Stochastic processes
Filtration Stopping time References
Stochastic processes
De…nition
De…nition
Let(Ω,F)be a measurable space. A stochastic process X = fXt :t 2 T g
is a family of random variables, all built on the same
measurable space(Ω,F) whereT represents a set of indices.
processes
Stochastic processes Filtration De…nitions Example
Stopping time References
Filtration I
De…nitions
De…nition
A familyF=fFt :t 2 T gof σ algebras on Ωis a …ltration on the measurable space(Ω,F) if
(F1) 8t 2 T,Ft F,
(F2) 8t1,t2 2 T such thatt1 t2,Ft1 Ft2.
De…nition
A stochastic processX =fXt :t 2 T g is said to be adapted to the …ltrationF=fFt :t2 T g if
8t 2 T,Xt is Ft measurable.
processes
Stochastic processes Filtration De…nitions Example
Stopping time References
Filtration II
De…nitions
De…nition
The …ltrationF=fFt :t 2 T g is said to be generated by the stochastic processX = fXt :t 2 T gif
8t 2 T,Ft =σfXs :s 2 T,s tg.
processes
Stochastic processes Filtration De…nitions Example
Stopping time References
Filtration I
Example
Example1. Let’s assume that the sample space is Ω=fω1,ω2,ω3,ω4gand that T =f0,1,2,3g. The stochastic processX = fXt :t 2 f0,1,2,3gg represents the evolution of a stock price,Xt = the stock price at close of market on thet th day, while time t =0 represents today.
ω X0(ω) X1(ω) X2(ω) X3(ω) ω1 1 0,50 1 0,50 ω2 1 0,50 1 0,50
ω3 1 2 1 1
ω4 1 2 2 2
processes
Stochastic processes Filtration De…nitions Example
Stopping time References
Filtration II
Example
Question. What is the …ltration generated by this stochastic process?
processes
Stochastic processes Filtration De…nitions Example
Stopping time References
Filtration III
Example
ω X0(ω) X1(ω) X2(ω) X3(ω) ω1 1 0,50 1 0,50 ω2 1 0,50 1 0,50
ω3 1 2 1 1
ω4 1 2 2 2
Answer.
F0 = σfX0g=f?,Ωg,
F1 = σfX0,X1g=σffω1,ω2g,fω3,ω4gg, F2 = σfX0,X1,X2g= σffω1,ω2g,fω3g,fω4gg, F3 = σfX0,X1,X2,X3g=σffω1,ω2g,fω3g,fω4gg.
Note that anyσ algebraF containing the sub-σ algebraF3 makeX0, X1,X2 andX3 F measurable.
processes
Stochastic processes Filtration De…nitions Example
Stopping time References
Filtration IV
Example
Recall that
F0 = σfX0g=f?,Ωg,
F1 = σfX0,X1g=σffω1,ω2g,fω3,ω4gg, F2 = σfX0,X1,X2g=σffω1,ω2g,fω3g,fω4gg, F3 = σfX0,X1,X2,X3g=σffω1,ω2g,fω3g,fω4gg. Interpretation.Ωrepresents states of nature. Xt(ωi)represents the stock price at timet if it is thei i th state of nature that has occurred. At time 0 (today), we know with certitude the stock price and we cannot identify which of the states of nature has occurred. That’s why the sub-σ algebra F0 is the trivialσ algebra, since it doesn’t contain any information.
processes
Stochastic processes Filtration De…nitions Example
Stopping time References
Filtration V
Example
Recall that
F1=σfX0,X1g=σffω1,ω2g,fω3,ω4gg.
1 At timet=1, we know a bit more. Indeed, if we observe a stock price of 0.50, then we know that the state of nature that has occurred is ω1 orω2 but certainly notω3 orω4. As a result, we can deduce that the stock price for the following two periods (t=2 and t=3) will be 1 and 0.50 dollar respectively.
2 On the contrary, if at timet=1, we observe a stock price of 2 dollars, then we know that the state of nature that has occurred is eitherω3 orω4. We can deduce from there that the stock price won’t fall back under the one-dollar level: because, after observing the process at timet=1, we’ll be able to determine whether event fω1,ω2gor eventfω3,ω4ghas happened,
F1=σffω1,ω2g,fω3,ω4gg.
processes
Stochastic processes Filtration De…nitions Example
Stopping time References
Filtration VI
Example
Recall that
F2=σfX0,X1,X2g=σffω1,ω2g,fω3g,fω4gg.
Let’s now assume that in timet=2, we observe a price of one dollar. A frequently made mistake is to conclude that the sub-σ algebra associated with that time isσffω1,ω2,ω3g,fω4gg since by observingX2 we are able to distinguish between the eventsfω1,ω2,ω3gandfω4g. That would be true if we were just beginning to observe the process, which is not the case. We must take into account the information obtained since timet=0. But the paths(X0(ω),X1(ω),X2(ω))enable us to distinguish between the three following events: fω1,ω2g,fω3gandfω4g. Indeed, after observing the prices until time two, we will know with certitude which state of natureω has occurred, unless we have observed path(1,12,1), in which case we’ll be unable to distinguish between states of natureω1 and ω2.
1Throughout this chapter, we go further into an example initiated in Stochastic Calculus, A Tool for Financeby Daniel Dufresne.
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time
Introduction
We will realize how very useful the concept of stopping time is when we will attempt to price American-style derivative
products. The main role of stopping times is to help determine the time when the option holder will exercise his or her right.
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time
De…nition
De…nition
Let(Ω,F)be a measurable space such that Card(Ω)< ∞ and equipped with the …ltrationF= fFt :t 2 f0,1, ...gg. A stopping time τis a (Ω,F) random variable that takes its values inf0,1, ...gand is such that
fω2 Ω:τ(ω) tg 2 Ft for all t 2 f0,1, ...g. (1) Exercise. Show that the condition (1) above is equivalent to
fω2 Ω:τ(ω) =tg 2 Ft for all t 2 f0,1, ...g.
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time I
Example
Example. Let’s return to the example described earlier: X represents a stock price.
ω X0(ω) X1(ω) X2(ω) X3(ω) ω1 1 0,50 1 0,50 ω2 1 0,50 1 0,50
ω3 1 2 1 1
ω4 1 2 2 2
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time II
Example
We had determined that the …ltration containing the information revealed by the process at each time is
F0 = σfX0g=f?,Ωg,
F1 = σfX0,X1g=σffω1,ω2g,fω3,ω4gg, F2 = σfX0,X1,X2g= σffω1,ω2g,fω3g,fω4gg, F3 = σfX0,X1,X2,X3g=σffω1,ω2g,fω3g,fω4gg.
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time III
Example
Recall that:
ω X0(ω) X1(ω) X2(ω) X3(ω)
ω1 1 0,50 1 0,50 ω2 1 0,50 1 0,50
ω3 1 2 1 1
ω4 1 2 2 2
We won’t sell our stocks today (t =0)but we will sell them as soon as the price is greater than or equal to 1.
The random time representing that situation is τ(ω1) =2, τ(ω2) =2, τ(ω3) =1 and τ(ω4) =1.
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time IV
Example
Such a random variable truly is a stopping time since
fω 2Ω:τ(ω) =0g = ?2 F0,
fω 2Ω:τ(ω) =1g = fω3,ω4g 2 F1, fω 2Ω:τ(ω) =2g = fω1,ω2g 2 F2, fω 2Ω:τ(ω) =3g = ?2 F3.
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time V
Example Recall that:
ω X0(ω) X1(ω) X2(ω) X3(ω)
ω1 1 0,50 1 0,50 ω2 1 0,50 1 0,50
ω3 1 2 1 1
ω4 1 2 2 2
Let’s now consider the random time τ modelling the following situation: we’ll buy stock as soon as it enables us to make a pro…t later.
Such a random value takes values τ (ω1) =1,
τ (ω2) =1, τ (ω3) =0 and τ (ω4) =0. τ is not a stopping time since
fω 2Ω:τ (ω) =0g=fω3,ω4g2 F/ 0.
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time VI
Example
Intuitively, the time τ when one makes a decision is a stopping time if the decision is made based on the
information available at that time. In the case of stopping times, using a crystal ball is prohibited.
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time I
Stopping time transformation
Theorem
Let(Ω,F), be a measurable space such that Card(Ω)<∞ and equipped with the …ltrationF= fFt :t 2 f0,1, ...gg. If the random variablesτ1 andτ2 are stopping times with respect to the …ltrationF, then τ1^τ2 minfτ1,τ2g and
τ1_τ2 maxfτ1,τ2gare also stopping times.
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time II
Stopping time transformation
Proof of the theorem. Ifτ is a stopping time, then 8t 2 f0,1, ...g
fω2Ω:τ(ω) tg 2 Ft. 8k 2 f0,1, ...g,
fω 2Ω:τ1(ω)^τ2(ω) kg
= fω 2Ω:τ1(ω) k or τ2(ω) kg
= f|ω 2Ω:τ{z1(ω) kg}
2Fk
[ f|ω2Ω:τ{z2(ω) kg}
2Fk
2 Fk.
Exercise. Prove the above result for τ1_τ2.
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time I
First passage time
De…nition
Let(Ω,F)be a measurable space such that Card(Ω)< ∞ and equipped with the …ltrationF= fFt :t 2 f0,1, ...gg. X = fXt :t 2 f0,1, ...gg represents a stochastic process adapted to that …ltration. LetB R a subset of the real numbers. We de…ne the time until the stochastic processX
…rst enters the setB as
τB(ω) =minft 2 f0,1, ...g:Xt(ω)2Bg. If it happened that the patht !Xt(ω)never hits the set B then we de…neτB(ω) =∞.
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time II
First passage time
Theorem
The random variableτB is a stopping time.
Proof of the theorem. Since Card(Ω)<∞, then 8t 2 f0,1, ...g,Xt can only take a …nite number of values.
Let’s denote them by
x1(t) <...<xm(tt). 8t 2 f0,1, ...g,
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time III
First passage time
fω2Ω:τB(ω) =tg
= fω2Ω:X0(ω)2/B, ...,Xt 1(ω)2/B,Xt(ω)2Bg
=
t\1 k=0
fω2Ω:Xk(ω)2/Bg
!
\ fω2Ω:Xt(ω)2Bg
= 0 B@
t\1 k=0
[ xi(k)/2B
n
ω2Ω:Xk(ω) =xi(k)o 1 CA
\ 0 B@ [
xi(t)2B
n
ω2Ω:Xt(ω) =xi(t)o1 CA
2 Ft
processes
Stochastic processes Filtration Stopping time
De…nition Example Transformations First passage References
Stopping time IV
First passage time
since
t\1 k=0
[ xi(k)2/B
n
ω2Ω:Xk(ω) =xi(k)o
| {z }
2FksinceXis adapted.
| {z }
2Fk FtsinceFkis aσ algebra
| {z }
2FtsinceFtis aσ algebra.
\ [
xi(t)2B
n
ω2Ω:Xt(ω) =xi(t)o
| {z }
2FtsinceXtisFt measurable.
| {z }
2FtsinceFtis aσ algebra.
processes
Stochastic processes Filtration Stopping time References
References
BILLINGSLEY, Patrick (1986). Probability and Measure, Second Edition, Wiley, New York.
DUFRESNE, Daniel (1996). Stochastic Calculus, A Tool for Finance, Department of Mathematics and Statistics, Université de Montréal.
KARLIN Samuel and TAYLOR Howard M. (1975). A First Course in Stochastic Processes, Second Edition, Academic Press, New York.