Given a filtered probability space $( \Omega, \mathcal{F}, (\mathcal{F}_t)_t, \mathbb{P})$ and an $\mathcal{F} \otimes \mathcal{B}(\mathbb{R}_+)$-measurable bounded process $X: \Omega \times \mathbb{R}_+ \rightarrow \mathbb{R}$, we know by the predictable projection theorem that there exists a process $^pX$ (unique up to evanescence) such that

  1. $^pX$ is predictable
  2. $(^pX)_T = \mathbb{E}(X_T | \mathcal{F}_{T-})$ on $\{T<\infty\}$ for every predictable stopping time $T$

Taking for granted this result (I am not sure whether of not it requires any assumption on the filtration), my question is about the predictable projection of a stopped process.

In particular, take $X$ measurable as in the theorem and consider the stopped process $X^S$, where $S$ is any optional time. How can we express $^p(X^S)$ in terms of $^pX$?

As far as I understand (Jacod & Shiryaev, page 23), it is the case that

\begin{equation} ^p(X^S) = (^pX) \mathbb{I}_{[\![0, S]\!]} + X_S \mathbb{I}_{]\!]S, \infty ]\!]} \tag{*} \end{equation}

but I cannot prove either of the defining properties.

With regard to (1), I was hoping to show that both terms on the rhs of (*) are predictable.

$(^pX) \mathbb{I}_{[\![0, S]\!]}$ is predictable, because it is the product of two predictable processes.

The problem is that I do not see why $X_S \mathbb{I}_{]\!]S, \infty ]\!]}$ needs to be predictable. In order to have $Y \mathbb{I}_{]\!]S, \infty ]\!]}$ predictable if $S$ is optional I need $Y$ to be $\mathcal{F}_S$-measurable, but $X_S$ is not necessarily $\mathcal{F}_S$-measurable.

What is even more puzzling is that - if I interpret Jacod & Shiryaev correctly - we can deduct from (*) that $X_S \mathbb{I}_{]\!]S, \infty ]\!]}$ is predictable (since it is the difference of two predictable processes) for every bounded measurable process $X$ and optional time $S$.

Would not $X(\omega, t) = \mathbb{I}_A(\omega)$, with, say, $A \in \mathcal{F} \setminus \mathcal{F}_1$ and $S(\omega) = \frac12$ be a trivial counterexample to this (in this example $X_S \mathbb{I}_{]\!]S, \infty ]\!]}$ would not even be adapted)?

With regard to (2), I am even more clueless.

Apologies for the macroscopic error that I am somewhere certainly making.


1 Answer 1


It seems to me that there is an error in formula 2.29 on page 23 of J&S. (Amazingly, the error persists in the second edition, which appeared 23 years after the first.) The formula is correct provided $X$ is optional. More generally, it would be correct if $X$ were replaced in the second term on the right side by the optional projection $\,{}^o\!X$ of $X$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.