Bivariate GBM - crosscovariance

131 Views Asked by At

I have troubles concerning a correlated bivariate GBM with identical drift and diffusion rates.

Let $dX^i_t = \mu X^i_t dt + \sigma X^i_tdW^i_t$

and $E[dW_t ^idW^j_t] = \rho_{i,j}dt$

If $X_0^i = 1$, I already learned that for $s<t$

$Cov(X^i_s, X^i_t) = exp((\mu+0.5\sigma^2)\cdot(s+t)) \cdot (exp(\sigma^2 s)-1)$

and

$Cov(X^i_t, X^j_t) = exp(2\mu t)\cdot(exp(\sigma^2\rho t)-1)$

The former may be referred to as time series covariance and the latter as cross sectional covariance.

However for $s<t$, I'm stuck on the quantity:

$Cov(X^i_s, X^j_t) = ???$

I appreciate your comments!

1

There are 1 best solutions below

0
On

I think I got it:

Since $Cov(X^i_s, X^j_t) = Cov[X^i_s, X^j_s\cdot exp((\mu-\sigma^2/2)(t-s)+\sigma(W(t)-W(s)))] $

it follows that

$Cov(X^i_s, X^j_t) = exp((\mu-\sigma^2/2)(t-s))\cdot Cov[X^i_s, X^j_s\cdot exp(\sigma W(t)-\sigma W(s))]$

and this yields

$Cov(X^i_s, X^j_t) = exp((\mu-\sigma^2/2)(t-s))\cdot Cov[X^i_s, X^j_s]$