A specific local martingale

157 Views Asked by At

Knowing that M and N are two local martingales. Can I do the following simple reasoning for claiming that $N^\tau(M^\tau - M)$ is a local martingale?

: Since a stopped local martingale is still a local martingale, and since the product of two local martingale is still a local martingale, then $N^\tau M^\tau$. Applying the same reasoning also at $N^\tau M$ I could conclude that it is also a local martingale. Finally the difference of 2 local martingale is still a local martingale.

It follows that $N^\tau(M^\tau - M)$ is a local martingale.

If this reasoning does not hold, what is wrong? How could I prove it without using Ito or advanced techniques?

2

There are 2 best solutions below

0
On BEST ANSWER

Let $M^{(n)}$ and $N^{(n)}$ denote $M$ and $N$ stopped at $\sigma_n\wedge\tau_n$. These are both bounded martingales. Form $X^{(n)}_t:=N^{(n)}_{t\wedge\tau}(M^{(n)}_{t\wedge\tau}-M^{(n)}_t)$. As previously noted (now corrected), $$ X^{(n)}_t=\cases{0,&$0\le t<\tau$,\cr N^{(n)}_\tau(M^{(n)}_{\tau}-M^{(n)}_t),&$t\ge \tau$.\cr} $$ Now decompose (for $0\le s<t$) $$ X^{(n)}_t=(1_{\{s<t\le\tau\}}+1_{\{s<\tau\le t\}} +1_{\{\tau\le s<t\}})\cdot X^{(n)}_t $$ into three terms. The conditional expectation of the first term (given $\mathcal F_s$) is trivial; for the second use the tower property of conditional expectations, conditioning first on $\mathcal F_\tau$; the third is easy because $M^{(n)}$ is a martingale and $N^{(n)}_\tau\cdot 1_{\{\tau\le s<t\}}$ is $\mathcal F_s$-measurable. It follows that $X^{(n)}$ is a martingale (for each $n$), and then that $X$ is a local martingale.

3
On

Observe that your product of local martingales, call it $X$, is equal to $$ X_t=\cases{0,&$0\le t<\tau$,\cr N_\tau(M_{\tau}-M_t),&$t\ge \tau$,\cr} $$ If, for example, $M$ and $N$ are continuous local martingales (or, more generally, locally bounded local martingales) then by localizing you can reduce to the case in which both $M$ and $N$ are bounded martingales. In this case it is clear that $X$ is integrable, and it is straightforward to compute $\Bbb E[X_t|\mathcal F_s]$ for $0\le s<t$, to see that $X$ is a martingale.