I am trying to understand the concept of strong Markov property quoted from Wikipedia:

Suppose that $X=(X_t:t\geq 0)$ is a stochastic process on a probability space $(\Omega,\mathcal{F},\mathbb{P})$ with natural filtration $\{\mathcal{F}\}_{t\geq 0}$. Then $X$ is said to have the strong Markov property if, for each stopping time $\tau$, conditioned on the event $\{\tau < \infty\}$, the process $X_{\tau + \cdot}$ (which maybe needs to be defined) is independent from $\mathcal{F}_{\tau}:=\{A \in \mathcal{F}: \tau \cap A \in \mathcal{F}_t ,\, \ t \geq 0\}$ and $X_{\tau + t} − X_{\tau}$ has the same distribution as $X_t$ for each $t \geq 0$.

Here are some questions that make me stuck:

  1. In $\mathcal{F}_{\tau}:=\{A \in \mathcal{F}: \tau \cap A \in \mathcal{F}_t ,\, \ t \geq 0\} $, what does $\tau \cap A $ mean? $\tau$ is a stopping time and therefore a random variable and $A$ is a $\mathcal{F}$-measurable subset, but what does $\tau \cap A$ mean?
  2. How is the process $X_{\tau + \cdot}$ defined from the process $X_{\cdot}$ ? Is it the translated version of the latter by $\tau$?
  3. How is the conditional independence between a process, such as $X_{\tau + \cdot}$, and the sigma algebra, such as $\mathcal{F}_{\tau}$, given an event, such as $\{\tau < \infty\}$, defined?

    Related question, is independence between a random variable and a sigma algebra defined as independence between the sigma algebra of the random variable and the sigma algebra?

  4. Is "$X_{\tau+ t} − X_{\tau}$ has the same distribution as $X_t$ for each $t \geq 0$" also conditional on the event $\{\tau < \infty\}$?

Thanks and regards!

  • 2
    $\begingroup$ After reading that I think, really, you should be learning about the strong Markov property from somewhere other than Wikipedia. $\endgroup$ – George Lowther Mar 9 '11 at 23:00
  • $\begingroup$ @George: I tried to look up in several books, many of which are just talking about the theorem of strong Markov property for Brownian motion. One (Lamperti's book) uses some notation that I don't understand yet. Maybe I will sometime but who knows. So I very much appreciate some nice and clear references to clear things up. $\endgroup$ – Tim Mar 9 '11 at 23:14
  • $\begingroup$ I don't have time to write up a proper answer now, but I do have a definition of the strong Markov property on my blog (but it assumes a good knowledge of measure-theoretic probability theory) almostsure.wordpress.com/2010/07/19/… $\endgroup$ – George Lowther Mar 9 '11 at 23:50
  • $\begingroup$ And in that Wikipedia page, they don't say what notation they are using for the bits you labeled (1) and (2), and it looks weird. (4) is just plain wrong. $\endgroup$ – George Lowther Mar 9 '11 at 23:54
  • $\begingroup$ @George: Thanks for your blog! In the third paragraph there, "Recall that the law of an inhomogeneous Markov process $X$ is described by a transition function ${\{P_t\}_{t\ge0}}$ on some measurable space ${(E,\mathcal{E})}$." Do you mean "inhomogeneous" or "homogenenous"? $\endgroup$ – Tim Mar 10 '11 at 10:37

Here is a less garbled version of the Wikipedia definition. (Use TheBridge's correction for the definition of ${\cal F}_\tau$.) The post-$\tau$ process $X_{\tau+\cdot}$ is defined on the event $\{\tau<\infty\}$ by $$ X_{\tau+t}(\omega) = X_{\tau(\omega)+t}(\omega),\qquad t\ge 0, $$ for $\omega\in\{\tau<\infty\}$. One way to state the strong Markov property is this: The conditional distribution of $X_{\tau+\cdot}$ given ${\cal F}_\tau$ is (a.s.) equal to the conditional distribution of $X_{\tau+\cdot}$ given $\sigma\{X_\tau\}$, on the event $\{\tau<\infty\}$. More precisely, $$ P[ X_{\tau+t}\in B|{\cal F}_\tau] = P[ X_{\tau+t}\in B|X_\tau],\qquad \hbox{almost surely on }\{\tau<\infty\}, $$ for all $t\ge 0$, and all measurable subsets $B$ of the state space of $X$.

This is equivalent to the statement that $X_{\tau+\cdot}$ and ${\cal F}_\tau$ are conditionally independent, given $X_\tau$: $$ P[ F\cap \{X_{\tau+t}\in B\}|X_\tau] = P[ F|X_\tau]\cdot P[X_{\tau+t}\in B|X_\tau],\qquad \hbox{almost surely on }\{\tau<\infty\}, $$

  • $\begingroup$ The condition you wrote after your "More precisely" characterizes the conditional distribution of the random variable $X_{\tau+t}$ conditionally on $\mathcal{F}_\tau$, for every $t\ge0$. This is strictly weaker than the property of the conditional distribution of the process $X_{\tau+\cdot}$ conditionally on $\mathcal{F}_\tau$ that is required for $X$ to be strong Markov. $\endgroup$ – Did Mar 10 '11 at 20:56
  • $\begingroup$ Good point! That really should be $$ P[ G|{\cal F}_\tau] = P[ G|X_\tau],\qquad \hbox{almost surely on }\{\tau<\infty\},\leqno(*) $$ for all events $G$ in the $\sigma$-field generated by $X_{\tau+\cdot}$, and similarly for the conditional independence form of the property. On the other hand, if the weaker form found in my first post holds {\it for all\/} stopping times, then so does the stronger form ($*$). $\endgroup$ – JohnD Mar 10 '11 at 21:34
  • $\begingroup$ D Sure, by the chain rule for conditional probabilities. But my point is that, since the OP asks for clear and explicit definitions and proofs about a notion he/she is trying to understand in depth, you might want to add to your post an expanded version of the remark you made in the last sentence of your comment. $\endgroup$ – Did Mar 11 '11 at 6:07

For the first one the notation is wrong it should I think $A\cap\{\tau\le t\}\in\mathcal{F}_t$ instead of $A\cap\tau$.

And for the fourth point, look at George Lowther comments below, that fully address the problematic.


  • $\begingroup$ No, the fourth point is saying that it has identical increments, which is true for Levy processes, but not part of the strong Markov property. $\endgroup$ – George Lowther Mar 9 '11 at 23:58
  • $\begingroup$ In fact, the fourth point on its own implies that it is a Levy process starting from zero. $\endgroup$ – George Lowther Mar 10 '11 at 0:03
  • $\begingroup$ You are right how I missed that I edit my post accordingly $\endgroup$ – TheBridge Mar 10 '11 at 10:53
  • $\begingroup$ I'm guessing, but I think the person who wrote that Wikipedia entry copied a statement from somewhere which was only dealing with Levy processes, and also made some typos. $\endgroup$ – George Lowther Mar 10 '11 at 12:18

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.