I want to show the following:

Let $W_t$ be a 1 dimensions brownian motion and $V_t= \int_{0}^{t} W_sds.$ Prove that the pair $(W_t,V_t)$ is a two-dimensional Markov process.

I know that the Brownian Motion alone is a Markov Process. The time integral of Brownian Motion is not a Markov Process, but I know it is normally distributed from these two questions: previous question 1 and previous question 2.

In order to show that $(W_t,V_t)$ is a Markov Process I know I need to verify that is $s,t \ge 0$ and then $P(X_{s+t}|F_s)=P(X_{s+t}|X_s)$ almost surely. How can I apply the information I know about $W_t$ alone being a Markov Process and $V_t$ being normally distributed to verify this fact?

  • $\begingroup$ Suggestion: Write $V_{t+s}=V_s+tW_s+\int_s^{t+s}(W_u-W_s)\,ds$. The integral on the right is independent of $\mathcal F_s$. $\endgroup$ – John Dawkins Apr 27 '17 at 1:31
  • $\begingroup$ @JohnDawkins How do I show the pair together make a markov process? Should I be taking $P(V_{s+t}|F_s)?$ Then by your comment $P(V_{s+t}|F_s)=P(V_s +tW_s + \int_{s}^{t+s}(W_u-W_s)ds|F_s)=P(\int_{s}^{t+s}(W_u-W_s)ds)+P(V_s +tW_s|F_s)$. How do I simplify this further? $\endgroup$ – user75514 Apr 27 '17 at 1:41

The suggestion lets one see that conditional on $\mathcal F_s$, the random pair $(W_{t+s},V_{t+s})$ has a bivariate normal distributed with mean vector $\left[\matrix{W_s\cr V_s + tW_s\cr}\right]$ and covariance matrix $\left[\matrix{ t&t^2/2\cr t^2/2&t^3/3}\right]$. Because this conditional distribution depends on $\mathcal F_s$ only through the pair $(W_s,V_s)$, we see that the pair process$\{(W_t,V_t): t\ge 0\}$ is Markovian.


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.