5
$\begingroup$

Let $B = (B_t)_{t\geq 0}$ be a standard brownian motion started at $0$. Consider the two following stochastic equations: \begin{equation} \begin{split} dX_t &=& (13 + 2X_t)\,dt + (6 + X_t)\,dB_t \\ dY_t &=& 2Y_t\,dt + Y_t\,dB_t \end{split} \end{equation} with $X_0 = Y_0 = 1$.

I am trying to show that the following identity in law holds: $$ X_t - Y_t\left(1 + 6\int_0^t\frac{1}{Y_s}\,dB_s \right) \stackrel{law}{=} 7\int_0^tY_s\,ds $$

Combining $X$ and $Y$, it can be shown through Ito's formula that the solution for $X$ is given by $$ X_t = Y_t\left(1 + 7\int_0^t\frac{1}{Y_s}\,ds + 6\int_0^t\frac{1}{Y_s}\,dB_s \right) $$

But I am having trouble operating with the first equation to work out the law identity, and the solution for $X$ is not really helping me. Any hint is most welcome, as I am a bit lost now. I suspect the solution might not be trivial.

$\endgroup$
  • 1
    $\begingroup$ So, basically your claim boils down to $$Y_t \int_0^t \frac{1}{Y_s} \, ds \stackrel{\text{law}}{=} \int_0^t Y_s \, ds, $$ right? $\endgroup$ – saz Feb 13 '15 at 11:56
  • $\begingroup$ yes, with a 7 multiplying the integral on the left of your identity also. Maybe now applying integration by parts? $\endgroup$ – Adam Feb 13 '15 at 13:27
3
$\begingroup$

Note that the solution of the SDE, $\mathrm dY_t = 2Y_t\mathrm dt + \mathrm dB_t\ (\mbox{for }Y_0 = 1)$, is a collection of log-normal random variables,

$$ Y_t = e^{\frac{3}{2}t + B_t}~(\mbox{for }t\geqslant 0)\,. $$

Therefore, for $0\leqslant s\leqslant t$,

$$ \left\{\frac{Y_t}{Y_s}\right\}_{0\leqslant s\leqslant t}\quad \stackrel{\text{law}}{=}\quad \left\{Y_{s}\right\}_{0\leqslant s\leqslant t}\,.\tag{1} $$

proof:

For $n\in\mathbb{N}$ and $0\leqslant t_1<t_2<\ldots<t_n\leqslant t$, the time-reversal properties of Brownian motion justify the following:

$ \,\,\,P\left(\frac{Y_t}{Y_{t_1}}\leqslant y_1,\ \ldots,\ \frac{Y_t}{Y_{t_n}}\leqslant y_n\right)\\ \begin{eqnarray*} &=& P\left(\frac{3}{2}(t-t_1) + B_t - B_{t_1}\leqslant \log y_1,\ \ldots,\ \frac{3}{2}(t-t_n) + B_t - B_{t_n}\leqslant \log y_n \right) \\ &=& P\left( B_t - B_{t_1}\leqslant \log y_1 - \frac{3}{2}(t-t_1),\ \ldots,\ B_t - B_{t_n}\leqslant \log y_n - \frac{3}{2}(t-t_n) \right) \\ &=& P\left( -(B_{t-r_1}-B_t)\leqslant \log y_1 - \frac{3}{2}r_1,\ \ldots,\ -(B_{t-r_n}-B_t)\leqslant \log y_n - \frac{3}{2}r_n \right) \\ &=& P\left( B_{t-r_1}-B_t\leqslant \log y_1 - \frac{3}{2}r_1, \ldots, B_{t-r_n}-B_t\leqslant \log y_n - \frac{3}{2}r_n \right) \\ &=& P\left( B_{r_1}\leqslant \log y_1 - \frac{3}{2}r_1,\ \ldots,\ B_{r_n}\leqslant \log y_n - \frac{3}{2}r_n \right) \\ &=& P\left( \frac{3}{2}r_1 + B_{r_1}\leqslant \log y_1,\ \ldots,\ \frac{3}{2}r_n + B_{r_n}\leqslant \log y_n \right) \\ &=& P\left(Y_{r_1}\leqslant y_1,\ \ldots,\ Y_{r_n}\leqslant y_n\right), \end{eqnarray*} $

where we defined $r_i:=t-t_i\ .$

The equality in law, $(1)$, implies $$ \int_0^t \frac{Y_t}{Y_s} \, ds \stackrel{\text{law}}{=} \int_0^t Y_s \, ds\,, $$

which, from your application of Ito's lemma and as deduced by @saz, is the relationship to be proved.

$\endgroup$

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.