Could an integral w.r.t a local martingale be a proper martingale?

128 Views Asked by At

The following is taken from "Stochastic differential equations and diffusion processes" by Ikeda & Watanabe.

Let $Y,X\in \mathcal M_2^{c,loc}$ (the space of continuous, square integrable, local-martingales).

If we define $\tilde{Y}$ by

$$ \tilde{Y}(t)=Y(t)-\langle Y,X\rangle_t,$$

Then $\tilde{Y}(t)$ is a local martingale.

Proof:

Assume first that $\tilde{Y}(t)$ is bounded in the sense that for each $t\geq 0$, $\tilde{Y}(t)\in \mathfrak L^{\infty}(\Omega)$.

By Ito's formula we have

$$d(M(t)\tilde{Y}(t))=\tilde{Y}(t)dM(t)+M(t)dY(t)$$

where $M(t)$ is the Doléans-Dade exponential of $X$ (we assume that $X$ satisfy the Novikov's condition and hence M(t) is a martingale).

At this point the author say that this implies that $M(t)\tilde{Y}(t)$ is a martingale, shouldn't it be a local-martingale?

We have that $$M(t)\tilde{Y}(t)=\int_0^t \tilde{Y}(t)dM(t)+\int_0^t M(t)dY(t)$$

We have that the first stochastic integral is a martingale since the integrand is bounded and the integrator is a square integrable martingale, now on the second stochastic integral we are integrating wrt a local martingale, I don't see how the martingale property could arise.

Do the author actually mean local-martingale or is there something I am not seeing?