I want to show the following result for continuous time submartingales:

Suppose $\{X_t\}_{t \ge0}$ is a right continuous submartingale with respect to the filtration $\{\mathcal{F}_t\}_{t \ge 0}$. I want to show that for stopping times $\sigma \leq \tau$, if EITHER

(1) $\tau \leq C < \infty$ or

(2) $\{X_t\}_{t \ge0}$ is uniformly integrable,

then $X_\sigma \leq E(X_\tau| \mathcal{F}_\sigma)$

My attempt/start of a proof:

It suffices to show that $E(X_\sigma) \leq E(X_\tau)$ for all stopping times $\sigma, \tau$ which satisfy either (1) or (2).

I know how to prove the discrete time theorem of the identical form, so if we define $\tau_n \equiv \inf\{k/2^n : k, n \in \mathbb{N}\, k/2^n \ge \tau\}$ and similarly for $\sigma_n$, we know that

$$X_{\sigma_n} \leq E(X_{\tau_n}| \mathcal{F}_{\sigma_n})$$

as it is easy enough to show that $\sigma_n, \tau_n$ are both (discrete time) stopping times.

Because $\mathcal{F}_\sigma \subseteq \mathcal{F}_{\sigma_n}$ for all $n$,

$$E(X_{\sigma_n}|\mathcal{F}_\sigma) \leq E(X_{\tau_n}| \mathcal{F}_{\sigma}) \quad \quad (1)$$ applying the tower property.

If I knew that $X_{\sigma_n}$ and $X_{\tau_n}$ were uniformly integrable and that $X_\sigma, X_\tau$ were both in $L^1$, I would be able to conclude the proof by taking expectations in (1) and subsequently taking limits because $X_{\tau_n} \xrightarrow{a.s.} X_\tau$ by right continuity and likewise for $\sigma_n$. Is the UI stuff true? I really don't know where to start from here. Any help would be massively appreciated.


Fix $k$. For $n\ge k$, one has $\sigma_n \le \tau_n \le \tau_k$, and so

$$ X_{\sigma_n} \le E(X_{\tau_k}|\mathcal F_{\sigma_n}) =:M_n.$$

Since $\{\mathcal F_{\sigma_n}\}_{n\ge1}$ is a decreasing sequence of $\sigma$-fields with intersection $\bigcap_{n \ge 1}\mathcal F_{\sigma_n} = \mathcal F_\sigma$, $\{M_n\}$ is a backwards martingale and hence converges a.s. and in $L^1$ to $E(X_{\tau_k}|\mathcal F_\sigma)$. Thus, since $X_{\sigma_n}\to X_{\sigma}$ a.s. one has

$$ X_\sigma \le E(X_{\tau_k}|\mathcal F_\sigma) $$

for all $k$. Notice now that $X_{\tau_k} \le E(X_{\tau_0}|F_{\tau_0})$, which is integrable (this follows by applying (1) or (2) to the discrete submartingale $\{X_n\}_n$). Since $X_{\tau_k}\to X_\tau$ a.s., it follows by Fatou's lemma that

$$ \limsup_{k\to\infty}E(X_{\tau_k}|\mathcal F_\sigma) \le E(X_\tau|\mathcal F_\sigma).$$

Hence $X_\sigma \le E(X_\tau | \mathcal F_\sigma)$, as claimed.


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.