5
$\begingroup$

Let B be a standard Brownian motion in 1 dimension. Define $\tau = \inf\{t\in[0,1]:B_t=\underset{s\in[0,1]}{\max}B_s\},$ and $\tau' = \sup\{t\in[0,1]:B_t=\underset{s\in[0,1]}{\max}B_s\}.$ Show that $\tau$ and $\tau'$ are not stopping times. My idea was to first assume that $\tau$ is a stopping time, then by strong Markov property $\{B_{\tau+t}-B_{\tau}:t \geq 0\}$ is a standard Brownian motion, and so the process fluctuates above and below $B_\tau$ at $\tau$, which contradicts the fact that $B_\tau$ is the maximum. I am wondering if I am heading towards the right direction.

$\endgroup$
3
  • $\begingroup$ Your guess is exactly right. If $\tau$ were a stopping time, by the strong Markov property $\beta_t:=B_{T+t}-B_T$ would be a standard Brownian motion, but then, a.s. $\beta_t\leq 0$, a contradiction. You can reach contradictions for $\tau'$ in the same spirit. $\endgroup$ – charlus Apr 1 at 2:53
  • 1
    $\begingroup$ But one has to show that $\tau < 1$ and $\tau' < 1$ a.s. in order to use the strong Markov property. It's easy for the inf since if $\tau =1$, then the maximum must be attained at 1, which is not possible. However, for sup, I am not sure how to rule out the possibility that there exists a sequence approaching 1 where the max is obtained on this sequence. $\endgroup$ – statfan Apr 1 at 3:32
  • $\begingroup$ That's a nice point. Maybe you can get out of it using some sort of time-reversal or time-inversion but I'm not sure. Regardless, I think you only need to verify $P(\tau'<1)>0$ to apply (a version of) the strong Markov property. $\endgroup$ – charlus Apr 1 at 14:18
3
$\begingroup$

As noted, if $\tau$ were a stopping time, then $\beta_t:=B_{t+\tau}-B_\tau$, $t\ge 0$, would be a standard Brownian motion. On the event $\{\tau<1\}$ you would then have $\beta_t\le 0$ for all sufficiently small $t$, which is inconsistent with the Brownian character of $\beta$. Therefore $P[\tau<1]$ would be $0$. But $$ {1\over 2} =P[B_1<0]\le P[\tau<1], $$ because $\max_{0\le s\le 1}B_s\ge 0$. Therefore $\tau$ cannot be a stopping time.

Likewise for $\tau'$. In fact, one can show that the time at which Brownian motion on the time inerval $[0,1]$ attains its maximum is unique; that is $P[\tau=\tau'<1]=1$.

$\endgroup$

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.