Deriving stochastic integral $ X+\frac{1}{2}\int_t^T Z_s^2 ds – \int_t^TZ_s dB_s$

brownian motionstochastic-calculusstochastic-integralsstochastic-processes

If $X,\eta<\infty$ where
$$ \exp(X) = \mathbb{E}[\exp(X)]+ \int_0^T \eta_s dB_s$$
Then let $$ \exp(Y_t) = \mathbb{E}[\exp(X)|\mathcal{F}_t]$$
Prove that for some $t\in [0,T]$, that there is a stochastic process $Z$ where the following holds true:
$$ Y_t = X+\frac{1}{2}\int_t^T Z_s^2 ds – \int_t^TZ_s dB_s$$

I know that $Y_t = \ln\bigl(\mathbb{E}[\exp(X)|\mathcal{F}_t]\bigr)$ . Then we define an Ito process $S_t = S_0 + \int_0^t \eta_s dB_s,$ such that $S_0 = \mathbb E[e^X]$, then $S_t = \mathbb E[\exp(X)|\mathcal{F}_t] >0,$ so $ln(S_t) = ln(\mathbb E[\exp(X)|\mathcal{F}_t])$, applying Ito's Lemma to this then gives
$$ ln(S_t) = ln(S_0) + \int_0^t f'(ln(S_s)) dX_s + \frac{1}{2} \int_0^t f''(ln(S_s)) d\langle ln(S)\rangle_s $$

I couldn't get the end answer from here though..

Best Answer

So from Itô (on $S_t$ and $ln$ to get $Y_t$'s SDE), you get the following : $$Y_t=Y_0 -\frac{1}{2} \int_0^t Z_s^2 ds + \int_0^t Z_sdB_s$$ with $Z_s= \frac{\eta_s}{e^{Y_s}}$.
Now observe the following : $$exp (Y_T)=exp (X)$$ so $$X=Y_T=Y_0 -\frac{1}{2} \int_0^T Z_s^2 ds + \int_0^T Z_sdB_s$$ (and by the way you can see that taking $t=0$, that $Y_0= ln (\mathbb E[exp(X)])$ but we won't need that).

We are almost done now take $Y_t +(X-X)$ and the following comes : $$Y_t=Y_t +(X-X)=X - (\frac{1}{2}\int_0^t Z_s^2 ds - \frac{1}{2}\int_0^T Z_s^2 ds) +(\int_0^t Z_sdB_s-\int_0^T Z_sdB_s)$$ or $$Y_t= X + \frac{1}{2}\int_0^T Z_s^2 ds -\int_t^T Z_sdB_s$$

So I think there is typo in your exercise as I have found a minus term on the anticipative term of Brownian integral in the final representation of $Y$ unless mistaken.

Related Question