Chg

Stochatic Processes

in

Financial Modeling
 Ānancidy
i
nunhi

Stock price
Interest rates Stochastic

Foreign Exchanges Processes

Market Index continuous time

e.g HS Index
 Chara
o .cm

ti nous maybe
Ysj
Non_differentiable hard 九 predict

short
Trend in long run difficult in
run
QHwtfindmatnd.rs
A Many choices

Gaussian Process

Markov Process

Martingale

similar Process
Self
Difsion
Levy Process
___ ___
Q Which one 如 choose

Criterion
Simple

Not too simpler

 Motion CBM
A Brownian

or
processes
derived from 吵
13
 Why

all complicated
models
Simplest among

It's everything

Gaussian Markov

BM

Self
similar Levy
Interpretabig fan Physics viewpoint
Random Walk
BM ⼆

Beautiful Properties

in Practice
Adequate
extension fractional BM
If not
Chapter 1

Stochastic processes

Stock prices move with time t, i.e., they are stochastic processes. However, there are so many
di↵erent processes to choose from. We will take a quick glance at these processes, each of which
possesses interesting properties.

Which one then should we choose to model stochastic processes? Ideally, it should be,
on one hand, sophisticated enough to capture the characteristics of stock prices, and on the
other hand, it should be simple enough to tractable mathematically. Which would be an ideal
candidate then?

Definition

A stochastic process on a probability space (⌦, F, P ) is a collection of r.v.s

{Xt , t 2 T } = {Xt (!), t 2 T, ! 2 ⌦}.

1. X is a continuous-time process if T = [0, a] (a could be 1).

X is a discrete-time process if T = 0, 1, 2, ... (aka time series)

2. X is a function of t and !.
(i) For fixed t, Xt (!) is a r.v.
(ii) For fixed !, Xt (!) is a realization, trajectory, or sample path.

Special classes of stochastic processes

1. Gaussian processes
X is a Gaussian process if (Xt1 , ..., Xtn ) is multivariate Gaussian for all n 1.

1
2. Markov processes.
X is an Markov process if P (A|Ft ) = P (A|Xt ) for all t, where Ft = (Xs , s  t).
So past and future are conditionally independent given the present.

3. Di↵usion.
A di↵usion is a path-continuous Markov process such that for all t and x:

µ(t, x) =: lim h 1 E[(Xt+h Xt )|Xt = x]

h!0+
2
(t, x) =: lim h 1 E[(Xt+h Xt hµ(t, Xt ))2 |Xt = x].
h!0+

A simple example of di↵usion is Xt = µt + Bt .

4. Levy processes
X is a Levy process if X0 = 0 a.s., and

(a) It has stationary and independent increments, i.e.,

• (Xt1 Xt0 , ..., Xtn Xtn 1 ) =d (Xt1 +h Xt0 +h , ..., Xtn +h Xtn 1 +h ) for all n and
h.
• Xt2 Xt1 , Xt3 Xt2 , · · · , Xtn Xtn 1 are independent for all t1 < ... < tn .
(b) X is stochastically continuous, if Xt !p Xs as t ! s for all s 0. That is,
limt!s P (|Xt Xs | > a) = 0.

We make several remarks.

• Some examples of Levy processes are:

– Brownian motion: Xt Xs ⇠ N (0, t s);
– Poisson process: Xt Xs ⇠ P oisson( (t s)).
– Cauchy process: Xt Xs ⇠ Cauchy.
– ......
• BM is the only continuous Levy process.
• Poisson process is not continous, but stochastically continuous. To see that, for
0 < a < 1,
(t s)
P (|Xt Xs | > a) = 1 P (|Xt Xs | = 0) = 1 e ! 0.

• Condition (b) serves to exclude processes with jumps at fixed (non-random) times,
which can be regarded as “calendar e↵ects” and are not interesting for our purposes.
It means that for a given time t, the probability of seeing a jump at t is zero:
discontinuities occur at random times.

5. Martingales
X is a martingale w.r.t. {Ft : t 0} if

(i) X is adapted to {Ft : t 0}, i.e., Xt is Ft -measurable,

(ii) E|Xt | < 1 for all t < 1,
(iii) E[Xt |Fs ] = Xs for all 0  s  t.

2
6. Self-similar processes
X is H-self-similar with Hurst index H > 0, i.e.,

{Xct1 , ..., Xctn } =d {cH Xt1 , ..., cH Xtn }, for all c > 0, ti ’s, and n 1. (0.1)