3
$\begingroup$

I have the following question as stated in the title:

For a stopping time $\tau $ does $E[(\tau \wedge t)1_A]=E[(\tau \wedge s)1_A] $ for $s \le t $ and any $A \in \mathcal F_{\tau \wedge s }$?

Here $\mathcal F_{\tau \wedge s } $ is the $\sigma $-algebra of $\tau \wedge s $-past.


The context is that I'm trying to show that $(B^2_{\tau \wedge t } - \tau \wedge t,\mathcal F_{\tau \wedge t })$ is a martingale and the question is here is a second step in showing that $E[\tau \wedge t|\mathcal F_{\tau \wedge s} ]=\tau \wedge s$.

Thanks in advance!

$\endgroup$
  • 2
    $\begingroup$ Consider the case where $A=\Omega$ and $\tau$ is a constant stopping time with the value $\tau_0$. Then certainly we should not expect $\tau_0\wedge t=E[\tau\wedge t]=E[\tau\wedge s]=\tau_0\wedge s$ to hold in general. $\endgroup$ – Sangchul Lee Sep 25 '19 at 17:04
2
$\begingroup$

We have \begin{alignat*}{2} \mathbb{E}[(\tau \wedge t)\mathbf{1}_A] & = \mathbb{E}\Big[\mathbb{E}\left[(\tau \wedge t)\mathbf{1}_A\;|\;\mathcal F_{\tau \wedge s }\right]\Big]\quad\text{by the tower property}\\ & = \mathbb{E}\Big[\mathbb{E}\left[(\tau \wedge t)\;|\;\mathcal F_{\tau \wedge s }\right]\mathbf{1}_A\Big]\quad\text{by measurability}\\ & \neq \mathbb{E}\Big[\mathbb{E}\left[(\tau \wedge s)\;|\;\mathcal F_{\tau \wedge s }\right]\mathbf{1}_A\Big]\\ & = \mathbb{E}\left[(\tau \wedge s)\mathbf{1}_A\right], \end{alignat*} so answer seems no. Counterexample: $\tau=\infty$ almost surely.


If $B$ is a standard Brownian motion, then $\mathbb E\left[ B^2_{t}\right] = {t}$ and $\mathbb E\left[ B^2_{t}\;\;|\;\; \mathcal{F}_s\right] = B_s^2 + {t-s}$.

What you need to show seems $$ \mathbb E\left[ B_{\tau\wedge t}^2 - {\tau\wedge t} \;\;|\;\; \mathcal{F}_{\tau\wedge s}\right] = B_{\tau\wedge s}^2 - {\tau\wedge s} $$ for which you need $(B_t-B_s)\amalg\mathcal F_s$ and $(B_t-B_s)\sim B_{t-s}$.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ What about if $\tau <\infty$ a.s.? $\endgroup$ – MrFranzén Sep 25 '19 at 16:34
  • $\begingroup$ If $\tau>t$ a.s., it still doesn't work at least. But here it seems you have a BM? $\endgroup$ – thomasb Sep 25 '19 at 16:37
  • $\begingroup$ Thank you for your answer. I realize now that I took the wrong approach. Of course I should apply the optional stopping theorem to the continous martingale $B_t-t $ and not $B_t $ and then trying to show the matringale property separately for $\tau \wedge t $. $\endgroup$ – MrFranzén Sep 25 '19 at 17:53

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.