If an SDE is driven by a Brownian motion, do their filtrations contain the same information?

56 Views Asked by At

I dont know where to start here and I cant find anything about it online. To me it seems that is should be true that an stochasic differential equation that driven by a Browinan motion generetes the same filatration, any uncertanity coming from the Browinan motion.

For simplicity we can consider a stochasic differential equation of the form

$dX_{t}=a(t,X_{t})dt+b(t,X_{t})dB_{t}$

where $B_{t}$ is the Brownian motion

1

There are 1 best solutions below

0
On BEST ANSWER

This is indeed a subtle point. We follow Karatzas-Shreve section 5.3 "Weak solutions".

By definition (2.1) a strong solution $X_{t}$ is adapted with respect to

$$\mathcal{F}_{t}:=\sigma(\sigma(\xi)\vee \mathcal{F}^{B}_{t}\vee \mathcal{N} ),$$

where $\xi$ is the initial condition and $\mathcal{N}$ is the collection of null sets in $\mathcal{F}^{B}_{\infty}$.

In the case of weak solution $(X,W)$, we first fix some filtration $\mathcal{F}_{t}$ and then ask that both $X$ and the driving BM $W$ are adapted to it. That allows for the possibility of not equaling the above augmentation:

$$\mathcal{F}_{t}\supsetneq \sigma(\sigma(\xi)\vee \mathcal{F}^{W}_{t}\vee \mathcal{N} ).$$

A standard example is the SDE

$$X_{t}=\int_{0}^{t}sgn(X_{s})dW_{s}.$$

We will record the details. First we start with a Brownian motion $(X_{t},\mathcal{F}_{t}^{X})$. Then we let

$$W_{t}:=\int_{0}^{t}sgn(X_{s})dX_{s}.$$

This implies that $W_{t}$ is BM too with some filtration $\mathcal{F}_{t}^{W}\subseteq \mathcal{F}_{t}^{X}$. We will show that this is sharp

$$\mathcal{F}_{t}^{W}\subsetneq \mathcal{F}_{t}^{X}$$

by getting a contradiction. By Tanaka's formula

$$W_{t}:=\int_{0}^{t}sgn(X_{s})dX_{s}=|X_{t}|-2L_{t}^{X}(0)$$

$$=|X_{t}|-\lim_{\epsilon\to 0}\frac{1}{2\epsilon}meas\{0\leq s\leq t: |X_{s}|\leq \epsilon\},$$

and so

$$\mathcal{F}_{t}^{W}\subseteq \mathcal{F}_{t}^{|X|}.$$

So then we would get

$$\mathcal{F}_{t}^{X}\subseteq \mathcal{F}_{t}^{W}\subseteq\mathcal{F}_{t}^{|X|},$$

which is a contradiction because the absolute value removes the information about the sign.