0
$\begingroup$

We consider the stochastic system $$\frac{dS_t}{S_t}=-R_t\,dW_t,$$ with $$dR_t=-R_t\,dt-R_t\,dW_t, \quad R_0>0.$$ We conjecture, and would like to show that $$\mathbb{E}[S_t^2] = S_0^2\,\mathbb{E}\left[\exp\left(-2\int_0^t R_s\,dW_s-\int_0^t R_s^2\,ds\right)\right]=+\infty \quad \text{for all } t>0.$$ A natural idea is to localize the stochastic exponential by setting $$\tau_n:=\inf\left\{u\ge0:\int_0^u R_s^2\,ds\ge n\right\}\wedge t,$$ and to perform a Girsanov change of measure on $[0,\tau_n]$. Writing $$Z_t^{(n)} := \exp\left(-2\int_0^{t\wedge\tau_n}R_s\,dW_s-2\int_0^{t\wedge\tau_n}R_s^2\,ds\right),$$ we have $$S_{t\wedge\tau_n}^2 = S_0^2\,Z_t^{(n)}\exp\left(\int_0^{t\wedge\tau_n}R_s^2\,ds\right),$$ and $Z^{(n)}$ is a true martingale. Defining $\mathbb Q_n$ by $$\frac{d\mathbb Q_n}{d\mathbb P}=Z_t^{(n)},$$ the Brownian motion is shifted into $$\widetilde W_u^{(n)}:=W_u+2\int_0^{u\wedge\tau_n}R_s\,ds,$$ and the process $R$ solves on $[0,\tau_n]$ $$dR_u=(-R_u+2R_u^2)\,du-R_u\,d\widetilde W_u^{(n)}.$$ The extra quadratic drift $2R_u^2$ implies that the tilted dynamics explodes: immediate application of the Feller criteria, which in turn indicates that the stopped moments $\mathbb E[S_{t\wedge\tau_n}^2]$ should become arbitrarily large. However, the main difficulty is to remove the localization, since from $S_{t\wedge\tau_n}^2\to S_t^2$ almost surely one only gets $$\mathbb E[S_t^2]\le \liminf_{n\to\infty}\mathbb E[S_{t\wedge\tau_n}^2].$$

$\endgroup$

0

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.