Let $X_1, X_2, X_3, \ldots$ be i.i.d. random variables with zero mean and let $S_n := X_1 + \ldots + X_n$. Does $T := \inf\{n: S_n > 0\}$ always have infinite first moment?

In the trivial case, where $X_i = 0$, we have $T = \infty$. For random walk, in which $X_i = \pm 1$ with probability $1/2$, it can be shown that $\mathbb{E}[T]=\infty$. I am wondering if it is always the case that $\mathbb{E}[T]=\infty$.


The answer is yes. This follows indirectly from the optional stopping theorem. The relevant formulation of the theorem says the following. If $X_n$ is a martingale, the increments of $X_n$ are "conditionally bounded", and $\mathbb{E}(\tau) < \infty$, then $\mathbb{E}[X_\tau]=\mathbb{E}[X_0]$. "Conditionally bounded" means that $\mathbb{E}(|X_{n+1}-X_n| \, | \mathcal{F}_n) \leq C$ independent of $n$.

Here $S_n$ is a martingale because it is a sum of iid variables and its increments are conditionally bounded by $\mathbb{E}[|X_i|]$. So the first two hypotheses hold. Now by contraposition, since the conclusion fails, the third hypothesis must also fail. That is, since $\mathbb{E}[S_\tau] > 0$ while $\mathbb{E}[S_0] = 0$, we conclude that $\mathbb{E}(\tau)=\infty$.

  • $\begingroup$ I am not sure what you meant by 'on average you are moving to the right'. In the problem, we assume $E[X_i]=0$, which means the expected moving distance is $0$. Can you please elaborate on this? $\endgroup$ – Zilin J. Mar 5 '15 at 21:34
  • $\begingroup$ @roy_24601 Sorry, I missed the zero mean requirement. $\endgroup$ – Ian Mar 5 '15 at 21:36
  • $\begingroup$ @roy_24601 I've corrected my post now. $\endgroup$ – Ian Mar 5 '15 at 21:39
  • $\begingroup$ In my case, the increment in the martingale might not be bounded. The problem didn't assume anything on $X_i$ other than being mean zero. $\endgroup$ – Zilin J. Mar 5 '15 at 21:47
  • $\begingroup$ @roy_24601 Sorry, what you actually need is that they are "conditionally bounded", meaning that $\mathbb{E}(|X_{n+1}| | \mathcal{F}_n) \leq C$ for a fixed $C$ independent of $n$. I think this follows from integrability of the $X_n$. $\endgroup$ – Ian Mar 5 '15 at 21:49

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.