Let $X$ be a (cadlag) Levy process with a triplet $(\gamma, \sigma, \nu)$ and it stochastic exponential $\mathcal E$, which is the (cadlag) solution of SDE $d\mathcal E_t=\mathcal E_{t-}dX_t$, $\mathcal E_0=1$. It is known that the explicit formula of $\mathcal E$ is $$\mathcal E_t = e^{X_t - \frac{\sigma^2 t}{2}}\prod_{0\leqslant s \leqslant t}(1+\Delta X_s)e^{-\Delta X_s}$$ (e.g., see Protter, Stochastic Integration and Differential Equations 2005), where the jump $\Delta X_t:=X_t - X_{t-}$.

Assume that $\mathbb F=(\mathcal F_t)_{t\geqslant 0}$ is the augmented natural filtration of $X$, then $\mathbb F$ is right continuous. Since $X$ has strong Markov property, $(X_{\tau +t}-X_\tau)_{t\geqslant 0}$ is independent of $\mathcal F_\tau$ for any stopping time $\tau<\infty$ a.s.

My question is whether the process $(\dfrac{\mathcal E_{t+\tau}}{\mathcal E_\tau})_{t\geqslant 0}$ is independent of $\mathcal F_\tau$? Here we assume that $\Delta X_t \neq -1$ for all $t$ a.s. for making our formula is well-defined.

From the formula above, we only need to check that whether $\prod_{0<s\leqslant t}(1+\Delta X_{\tau +s})e^{-\Delta X_{\tau +s}}$ is independent from $\mathcal F_\tau$? By the strong Markov property of $X$, one can show that $\Delta X_{\tau + s}$ is independent from $\mathcal F_\tau$ for each $s\in (0,t]$. But I'm stuck to prove the same conclusion for the whole product. Does anybody have idea?


First of all, the number of jumps $N^{\tau}_t$ that happen between times $\tau$ and $t+\tau,$ the random times $t_i$ at which they happen and the random jumps $\Delta X_{\tau + t_i}$ are jointly independent of $\mathcal{F}_{\tau}$ by the strong Markov property.

Thus for any $A \in \mathcal{F}_{\tau}$ we see that for $Y_{t_i}$ defined according to your problem: $$ \mathbb{E}\bigg[1_A \prod_{i=1}^{N^{\tau}_t} Y_{\tau +t_i} \bigg] = \sum_k\mathbb{E}\bigg[1_A 1_{\{N^{\tau}_{t}=k\}} \prod_{i=1}^{k} Y_{\tau +t_i} \bigg] $$

which now conveniently factorizes due to the assumed inependence.

  • $\begingroup$ Can you explain more details why $N_t^\tau$, $t_i$, $\Delta X_{\tau +t_i}$ are jointly independent of $\mathcal F_\tau$? It is not clear for me. $\endgroup$ – ntt Jan 13 '18 at 13:31

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.