Let $\{X_n \}$ be a uniformly integrable martingale w.r.t. the natural filtration $\{ \mathcal{F}_n \}$.

Is $\{ X_\tau : \tau \text{ is a stopping time w.r.t. } \{ \mathcal{F}_n \} \}$ uniformly integrable?

I've tried using this theorem that conditional expectations of integrable random variables forms a uniformly integrable family, but this isn't working because I don't see how $X_\tau$ is actually the same thing as $\mathbb{E}(X_n | \mathcal{F}_\tau)$.

I'm really not sure how to proceed here... can anyone help?

Note that $X_\tau := \sum\limits_{n=0}^\infty X_n \textbf{1} \{ \tau = n\}$.

  • $\begingroup$ See Theorem 9.3.5 in Chung's 'A Course in Probability Theory'. $\endgroup$ – Kabo Murphy Sep 6 '18 at 8:41
  • $\begingroup$ You have to apply the theorem in Chung with $\tau$ as one stopping time and $\infty$ as the other. Define $X_{\infty}$ as the limit of the martingale (so $X_n=E(X|\mathcal F_n)$). $\endgroup$ – Kabo Murphy Sep 6 '18 at 9:08

Since the martingale $(X_n)_{n \in \mathbb{N}}$ is uniformly integrable, the limit $X_{\infty} = \lim_{n \to \infty} X_n$ exists pointwise and in $L^1$; in particular, $$\forall n \in \mathbb{N}_0: \quad \mathbb{E}(X_{\infty} \mid \mathcal{F}_n) = X_n. \tag{1}$$

Claim: For any stopping time $\tau$ it holds that $\mathbb{E}(X_{\infty} \mid \mathcal{F}_{\tau}) = X_{\tau}$.

Indeed: For $F \in \mathcal{F}_{\tau}$ we have $F \cap \{\tau=n\} \in \mathcal{F}_n$, and so

$$\int_F X_{\tau} \, d\mathbb{P} = \sum_{n \geq 0} \int_{F \cap \{\tau=n\}} \underbrace{X_{\tau}}_{=X_n} \, d\mathbb{P} \stackrel{(1)}{=} \int_{F \cap \{\tau=n\}} X_{\infty} \, d\mathbb{P} = \int_F X_{\infty} \, d\mathbb{P}.$$

Since $X_{\tau}$ is $\mathcal{F}_{\tau}$-measurable, this proves $\mathbb{E}(X_{\infty} \mid \mathcal{F}_{\tau}) = X_{\tau}$.

Applying the result mentioned in your question, we conclude that the family $\{X_{\tau}; \tau$ stopping time$\}$ is uniformly integrable.


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.