I was wondering how to prove/compute the differential of the product of two Brownian motions. I know how to do it in case they are independent as follows:

Suppose $dX_t= \mu_t dt +\sigma_t dW_t$ and $dY_t = v_t dt + \rho_t d \bar{W}_t$, where $W_t$ and $\bar{W}_t$ are independent, i.e. $(dW_t d\bar{W}_t= 0)$

Then I can prove $d(X_t Y_t)$ with the following trick:

I start with decomposition: $(X_t+Y_t)^2=X_t^2+Y_t^2+ 2X_t Y_t$,

Which leads by differentiation to $d(X_t Y_t)= \frac{1}{2}[d(\{X_t +Y_t\}^2) - d(X_t^2) - d(Y_t^2)]$

Then I apply Ito-lemma to all three parts separately as follows: $d(X_t^2)=(2\mu_tXt+\sigma_t^2)dt+ 2\sigma_t X_tdW_t= 2X_t dX_t + \sigma_t^2dt$ $d(Y_t^2)=(2v_t Y_t+\rho_t^2)dt+ 2\rho_t Y_t d\bar{W}_t= 2Y_t dY_t+\rho_t^2 dt$ $d((X_t+Y_t)^2) = 2(X_t+Y_t)dX_t+ 2(X_t+Y_t)dY_t + (\sigma_t^2+\rho_t^2)dt$

Combined I eventually find: $d(X_t Y_t) = Y_t dX_t + X_t dY_t$.

Now I was wondering how to do the prove in a similar fashion in the case of correlated brownian motions. Any help would be appreciated.


Well it's simpler to tackle it multidimensionnal Itô with $f(x,y)=x.y$, this gives you : $$d(XY)_t=f_{x}(X_t,Y_t)dX_t +f_{y}(X_t,Y_t)dY_t + \frac{1}{2} f_{xx}(X_t,Y_t)d<X>_t+\frac{1}{2}f_{yy}(X_t,Y_t)d<Y>_t +f_{xy}(X_t,Y_t)d<X,Y>_t$$

$$d(XY)_t=Y_t.dX_t +X_t.dY_t + \frac{1}{2}.0.d<X>_t+\frac{1}{2}.0.d<Y>_t +1.d<X,Y>_t$$ $$d(XY)_t=Y_t.dX_t +X_t.dY_t + d<X,Y>_t$$ Now $d<X,Y>_t=\sigma_t.\rho_td<B,\tilde{B}>_t=c.\sigma_t.\rho_t.dt$ where $d<B,\tilde{B}>=c.dt$ so that : $$d(XY)_t=Y_t.dX_t +X_t.dY_t + c.\sigma_t.\rho_t.dt$$

  • $\begingroup$ Thanks for the reply! What does the < > notation represent? I'm relatively new to stochastic processes. Would it also be possible to do it in a similar manner as my independence BM example? $\endgroup$
    – Rik
    Oct 19 '18 at 21:45
  • $\begingroup$ <> is for the square bracket process or if you would like in a continuous context the quadarituc variation and covariation of a process. Take a look at george lowther blog almost sure to understand it. I can be defined in at least two ways first the limit of quadartic variations which a construction "à la main" (almostsure.wordpress.com/2010/01/18/… ) or using the Doob-Meyer decomposition of square integrable martingales which is nice but more abstract.(see for example Karatzas and Shreve book) regards $\endgroup$
    – TheBridge
    Oct 19 '18 at 22:53

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.