Is $X$ adapted not sufficient for measurability of $X_T$ in $\mathcal F_T$ (the σ-algebra of a stopping time)?

82 Views Asked by At

Let $X$ be a stochastic process in continuous time and $T$ be a stopping time.

All the standard texts I've looked at so far, as well as this excellent blog, introduce the assumption that $X$ is progressively measurable to prove that $X_T$ is $\mathcal F_T$-measurable. Why is merely adapted $X$ not enough? Indeed the proof in the linked article seems to use "progressive" solely to claim "adapted" for the conclusion.

1

There are 1 best solutions below

2
On BEST ANSWER

The proof in the article uses progressive to claim that the stopped process $X^T$ is adapted, not that $X$ is adapted. The reason that something stronger than adaptedness is needed is because $X_T$ may fail to even be $\mathcal F_\infty$ measurable without further assumptions.

As an example, consider the probability space $\Omega := [0,1]$ equipped with the Lebesgue measure and the Borel (or Lebesgue) $\sigma$-algebra $\mathcal F$. We will use the constant filtration $\mathcal F_t = \mathcal F$ for all $t$. Define the stopping time $T$ by $T(\omega) = \omega$ for all $\omega \in \Omega$.

Let $A \subset [0,1]$ be any subset which is not $\mathcal F$ measurable, and define the stochastic process $X$ by $X_t(\omega) = 1_{\omega = t, t \in A}$. Observe that, for all $t$, $$\mathbb{P}({X_t \ne 0}) = \mathbb{P}(\omega = t, t \in A) \le \mathbb{P}(\omega = t) = 0, $$ so $X_t = 0$ almost surely. In particular $X_t$ is $\mathcal F_t$ measurable, i.e. $X$ is adapted. However, $$ X_{T(\omega)}(\omega) = 1_{\omega = T(\omega),T(\omega) \in A} = 1_{\omega = \omega, \omega \in A} = 1_{\omega \in A}, $$ which is not $\mathcal F$ measurable by the definition of $A$. Hence $X_T$ is not $\mathcal F$ measurable, and in particular is not $\mathcal F_T$ measurable.