2
$\begingroup$

Say I have the following two stochastic differential equations: $$\frac{dS_1}{S_1} = 0.03dt + 0.1 dW_1 + 0.4dW_2$$ $$\frac{dS_2}{S_2} = 0.03dt + 0.5 dW_1 + 0.1dW_2$$ where the instantaneous correlation between the two shocks are $cov(dW_1, dW_2)=0.5dt$. Let $S_1(0) = 100$ and $S_2(0) = 80$. Calculate $\mathbb{E}[1(S_1(1)<50)]$.

My attempt: The expectation is just simply the probability $\mathbb{P}(S_1(1)<50)$. However, I am confused with the extra $dW_2$ term in the stochastic differential equation for $S_1$. If the $dW_2$ was not there, for example, then we have a Geometric Brownian Motion (GBM) and we can use the properties of a GBM to calculate the required probability, but what do I do when there are two (correlated) Wiener process terms?

$\endgroup$
1
$\begingroup$

Generally, the respective solution to $$ {\rm d}S_t=S_t\left(\mu{\rm d}t+\sigma_1{\rm d}W_t^1+\sigma_2{\rm d}W_t^2\right) $$ with $$ {\rm d}W_t^1{\rm d}W_t^2=\rho{\rm d}t $$ would be $$ S_t=S_0\exp\left[\left(\mu-\frac{1}{2}\sigma_1^2-\frac{1}{2}\sigma_2^2-\rho\sigma_1\sigma_2\right)t+\sigma_1W_t^1+\sigma_2W_t^2\right]. $$

According to the SDE, the quadratic variation of $S_t$ is, intuitively, \begin{align} {\rm d}\left<S\right>_t&={\rm d}S_t{\rm d}S_t\\ &=S_t^2\left(\mu{\rm d}t+\sigma_1{\rm d}W_t^1+\sigma_2{\rm d}W_t^2\right)^2\\ &=S_t^2\left(\mu^2{\rm d}t^2+\sigma_1^2{\rm d}W_t^1{\rm d}W_t^1+\sigma_2^2{\rm d}W_t^2{\rm d}W_t^2+\right.\\ &\quad\quad\left.2\mu\sigma_1{\rm d}t{\rm d}W_t^1+2\mu\sigma_2{\rm d}t{\rm d}W_t^2+2\sigma_1\sigma_2{\rm d}W_t^1{\rm d}W_t^2\right)\\ &=S_t^2\left(\sigma_1^2{\rm d}t+\sigma_2^2{\rm d}t+2\sigma_1\sigma_2\rho{\rm d}t\right). \end{align}

Thus thanks to Ito's lemma, \begin{align} {\rm d}\log S_t&=\frac{{\rm d}S_t}{S_t}-\frac{1}{2}\frac{{\rm d}\left<S\right>_t}{S_t^2}\\ &=\left(\mu{\rm d}t+\sigma_1{\rm d}W_t^1+\sigma_2{\rm d}W_t^2\right)-\frac{1}{2}\left(\sigma_1^2+\sigma_2^2+2\rho\sigma_1\sigma_2\right){\rm d}t\\ &=\left(\mu-\frac{1}{2}\sigma_1^2-\frac{1}{2}\sigma_2^2-\rho\sigma_1\sigma_2\right){\rm d}t+\sigma_1{\rm d}W_t^1+\sigma_2{\rm d}W_t^2\\ &={\rm d}\left[\left(\mu-\frac{1}{2}\sigma_1^2-\frac{1}{2}\sigma_2^2-\rho\sigma_1\sigma_2\right)t+\sigma_1W_t^1+\sigma_2W_t^2\right]. \end{align} This eventually yields $$ S_t=S_0\exp\left[\left(\mu-\frac{1}{2}\sigma_1^2-\frac{1}{2}\sigma_2^2-\rho\sigma_1\sigma_2\right)t+\sigma_1W_t^1+\sigma_2W_t^2\right]. $$

With this solution, as well as the given correlation between $W_t^1$ and $W_t^2$, it is no longer hard to figure out the distribution of $S_t$, and the probability would follow straightforwardly.

$\endgroup$
  • $\begingroup$ Thank you! This is exactly what I was looking for. Could you take a look at another question of mine? math.stackexchange.com/questions/2758487/… $\endgroup$ – elbarto Apr 29 '18 at 8:06
  • $\begingroup$ @elbarto: My pleasure. I will have a look at this. $\endgroup$ – hypernova Apr 29 '18 at 8:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.