Progressive Measurability: Information on Each Interval
Adaptedness says that at each fixed time \(t\), the random variable \(X_t\) depends measurably only on the information \(\mathcal{F}_t\) available by that time. It says nothing by itself about how the whole collection of values \((X_s)_{0\le s\le t}\) varies jointly with time and outcome. Progressive measurability imposes that additional joint condition, separately on every finite time interval. It is useful, in particular, when processes are used as integrands in stochastic calculus.
Work on a filtered probability space \((\Omega,\mathcal{F},(\mathcal{F}_t)_{t\ge0},\mathbb{P})\), where \((\mathcal{F}_t)\) is increasing: \(\mathcal{F}_s\subseteq\mathcal{F}_t\) whenever \(s\le t\). Let \(X=(X_t)_{t\ge0}\) be a real-valued process. Write \(\mathcal{B}([0,T])\) for the Borel sigma-field on \([0,T]\).
The horizon \(T\) matters: the entire restricted path through time \(T\) must be jointly measurable using \(\mathcal{F}_T\). In particular, this is stronger than checking only that each separate time section is measurable. A process defined only through a finite time \(T_0\) can be called progressive on \([0,T_0]\) by checking all horizons \(T\le T_0\); to call it progressive on \([0,\infty)\), its values after \(T_0\) must also be specified and checked.
Progressive Measurability Implies Adaptedness
Proof. Fix \(t\ge0\). By progressive measurability, the restriction of \((s,\omega)\mapsto X_s(\omega)\) to \([0,t]\times\Omega\) is \(\mathcal{B}([0,t])\otimes\mathcal{F}_t\)-measurable. The map \(\omega\mapsto(t,\omega)\) is measurable from \((\Omega,\mathcal{F}_t)\) to \(([0,t]\times\Omega,\mathcal{B}([0,t])\otimes\mathcal{F}_t)\): the preimage of a measurable rectangle \(A\times C\) is \(C\) if \(t\in A\), and is empty otherwise. Measurability then holds for the sigma-field generated by these rectangles. Composing with the restricted process shows that \(\omega\mapsto X_t(\omega)\) is \(\mathcal{F}_t\)-measurable. Since \(t\) was arbitrary, \(X\) is adapted. \(\square\)
The proposition uses the full joint-measurability condition on \([0,t]\times\Omega\); it does not assert that adaptedness alone implies progressiveness. The examples below give useful situations in which the stronger property can be checked directly.
Finite Step Processes and Their Endpoints
Proof. Fix a horizon \(u\in[0,t_n]\). For each \(i<n\), let \(I_i=[t_i,t_{i+1})\cap[0,u]\). If \(I_i\) is nonempty, then \(t_i\le u\), so \(\xi_i\) is \(\mathcal{F}_{t_i}\)-measurable and hence \(\mathcal{F}_u\)-measurable by the increasing-filtration property. The map \((s,\omega)\mapsto \xi_i(\omega)\) is therefore \(\mathcal{B}([0,u])\otimes\mathcal{F}_u\)-measurable, and restricting it to the Borel time set \(I_i\) preserves measurability. When \(u=t_n\), the remaining singleton \(\{t_n\}\times\Omega\) carries the value \(\xi_n\), which is \(\mathcal{F}_{t_n}=\mathcal{F}_u\)-measurable. When \(u<t_n\), that singleton is not part of the restriction. The restriction to \([0,u]\times\Omega\) is thus measurable for every \(u\), as required. \(\square\)
Worked Example: A Step Process on a Finite Horizon
Take \(0<1<2\) and suppose \(\xi_0\) is \(\mathcal{F}_0\)-measurable, \(\xi_1\) is \(\mathcal{F}_1\)-measurable, and \(\xi_2\) is \(\mathcal{F}_2\)-measurable. On \([0,2]\), set \(X_s=\xi_0\) for \(0\le s<1\), \(X_s=\xi_1\) for \(1\le s<2\), and \(X_2=\xi_2\). Check progressiveness at horizons \(u=\tfrac12,1,\tfrac32,2\).
At \(u=\tfrac12\), the restriction is the constant-in-time value \(\xi_0\). Since \(\mathcal{F}_0\subseteq\mathcal{F}_{1/2}\), this restriction is measurable with respect to \(\mathcal{B}([0,\tfrac12])\otimes\mathcal{F}_{1/2}\).
At \(u=1\), the restriction uses \(\xi_0\) on \([0,1)\) and \(\xi_1\) at the singleton \(\{1\}\). Both coefficients are \(\mathcal{F}_1\)-measurable, and both time sets are Borel. At \(u=\tfrac32\), it uses \(\xi_0\) on \([0,1)\) and \(\xi_1\) on \([1,\tfrac32]\); both coefficients are \(\mathcal{F}_{3/2}\)-measurable. At \(u=2\), it uses \(\xi_0\) on \([0,1)\), \(\xi_1\) on \([1,2)\), and \(\xi_2\) at \(\{2\}\); each coefficient is measurable with respect to \(\mathcal{F}_2\).
Check: the value at \(s=1\) is \(\xi_1\), not \(\xi_0\), and the value at \(s=2\) is \(\xi_2\), not \(\xi_1\). These endpoint assignments are included in the restrictions just checked.
A Process Defined for All Times
For a process on \([0,\infty)\), a finite step description must include a continuation after its last listed time. The following example makes every boundary value explicit. Let \(B=(B_t)_{t\ge0}\) be Brownian motion adapted to \((\mathcal{F}_t)\), and define
Worked Example: Checking Every Horizon Boundary
Verify that this process is progressively measurable on \([0,\infty)\). Brownian adaptation gives \(B_1\in\mathcal{F}_1\) and \(B_2\in\mathcal{F}_2\), so \(\mathbf{1}_{\{B_1\ge0\}}\) is \(\mathcal{F}_1\)-measurable and \(B_2\) is \(\mathcal{F}_2\)-measurable. The constants \(2\) and \(0\) are measurable with respect to every sigma-field.
Fix any \(T\ge0\). On the time sets \([0,1)\cap[0,T]\), \([1,2)\cap[0,T]\), \([2,3)\cap[0,T]\), and \([3,\infty)\cap[0,T]\), the process has respectively the four displayed values. Whenever the second, third, or fourth time set is nonempty, its coefficient is measurable with respect to \(\mathcal{F}_T\): the second set requires \(T\ge1\), the third requires \(T\ge2\), and the fourth requires \(T\ge3\). Each time set is Borel, so the restriction is jointly measurable. This argument includes the boundary cases: for \(T=1\), the singleton \(\{1\}\) has value \(\mathbf{1}_{\{B_1\ge0\}}\); for \(T=2\), \(\{2\}\) has value \(B_2\); and for \(T=3\), \(\{3\}\) has value \(0\). Thus the condition holds for every \(T\).
Check: if \(B_1=-1\), then \(X_1=0\), whereas \(X_s=2\) for \(s<1\). The restriction at \(T=1\) is therefore not simply the constant \(2\) on all of \([0,1]\); it includes the distinct value at the endpoint.
The Continuous-Adapted Criterion
Proof. Fix \(T>0\). For each integer \(m\ge1\), put \(t_k^{(m)}=kT/2^m\) for \(k=0,\ldots,2^m\). Define \(X^{(m)}_s=X_{t_k^{(m)}}\) when \(t_k^{(m)}\le s<t_{k+1}^{(m)}\), and \(X^{(m)}_T=X_T\). For each \(k<2^m\), adaptation and the increasing filtration give \(X_{t_k^{(m)}}\) as an \(\mathcal{F}_T\)-measurable random variable; \(X_T\) is also \(\mathcal{F}_T\)-measurable. The time intervals and the endpoint singleton are Borel subsets of \([0,T]\). Consequently, each \(X^{(m)}\), viewed as a map on \([0,T]\times\Omega\), is \(\mathcal{B}([0,T])\otimes\mathcal{F}_T\)-measurable. For every \((s,\omega)\), the left grid point used in \(X^{(m)}_s(\omega)\) approaches \(s\) as \(m\to\infty\); at \(s=T\), the approximation is exactly \(X_T(\omega)\). Path continuity therefore gives \(X^{(m)}_s(\omega)\to X_s(\omega)\). A pointwise limit of measurable real-valued maps is measurable, so the restriction of \(X\) to \([0,T]\times\Omega\) is measurable. For \(T=0\), the restriction is the \(\mathcal{F}_0\)-measurable variable \(X_0\) on a singleton time interval. Hence the required condition holds for every \(T\ge0\). \(\square\)
Worked Example: A Continuous Adapted Brownian Process
Let \(X_t=B_t^2+t\). Brownian motion is adapted, so \(B_t\) is \(\mathcal{F}_t\)-measurable. The map \(x\mapsto x^2\) is Borel measurable, and \(t\) is deterministic; therefore \(X_t\) is \(\mathcal{F}_t\)-measurable for each fixed \(t\). Brownian paths are continuous, and the functions \(x\mapsto x^2\) and \(t\mapsto t\) are continuous, so \(X\) has continuous paths. The theorem gives progressive measurability.
For an explicit finite-horizon check, fix \(T>0\) and use the grids \(t_k^{(m)}=kT/2^m\). The step approximation on \([t_k^{(m)},t_{k+1}^{(m)})\) has coefficient \(B_{t_k^{(m)}}^2+t_k^{(m)}\), which is \(\mathcal{F}_{t_k^{(m)}}\)-measurable and hence \(\mathcal{F}_T\)-measurable. At \(T\), use the coefficient \(B_T^2+T\), which is \(\mathcal{F}_T\)-measurable. These measurable step maps converge pointwise to \(B_s^2+s\) on \([0,T]\times\Omega\) by path continuity, verifying the criterion on this horizon.
Check: at \(s=0\), \(X_0=B_0^2+0=0\). At the final time \(s=T\), the approximation uses \(B_T^2+T\) itself, so no endpoint is omitted.
Why the Distinction Matters
Progressiveness packages two kinds of measurability: the values at time \(s\) are compatible with information available by the horizon, and the process is jointly measurable across all times up to that horizon. The implication to adaptedness follows from taking a time section, but the reverse implication does not follow from that argument: knowing every section separately does not establish joint measurability of the restriction.
In stochastic calculus, integrands for the Ito integral are taken from the class \(\mathcal{L}^2\) of progressively measurable processes \(H\) satisfying \(\mathbb{E}\int_0^T H_s^2\,ds<\infty\) on the horizon under consideration. Progressive measurability is the measurability condition; the displayed square-integrability condition is a separate requirement. The continuous-adapted criterion supplies a common way to verify the former, while step-process checks make clear why time intervals and endpoint values must be handled precisely.
Common pitfall. A tempting but incorrect endpoint check for the global step example is: “At \(T=1\), only the first interval contributes, so \(X_s=2\) throughout \([0,1]\).” This treats \([0,1)\) as if it included \(1\). For the outcome \(B_1=-1\), the claimed value at \(s=1\) would be \(2\), but the definition gives \(X_1=\mathbf{1}_{\{-1\ge0\}}=0\). The corrected restriction has value \(2\) on \([0,1)\) and value \(\mathbf{1}_{\{B_1\ge0\}}\) on \(\{1\}\). The singleton is measurable at horizon \(1\), so the progressive check succeeds without misdescribing the process.
Check Your Understanding
For each question, state the relevant horizon and account for any endpoint that belongs to the restricted interval.
- Why does progressive measurability imply that \(X_t\) is \(\mathcal{F}_t\)-measurable for each fixed \(t\)?
- In the finite step-process proposition, which coefficient is used at \(t_i\), and which sigma-field makes it measurable?
- For the global Brownian step process, what values occur at \(T=1\), \(T=2\), and \(T=3\), respectively?
- In the proof for continuous adapted processes, why are the step-approximation coefficients measurable with respect to \(\mathcal{F}_T\)?
- For \(X_t=B_t^2+t\), identify separately the facts that establish adaptation and path continuity.
- What additional condition, apart from progressive measurability, appears in the definition of \(\mathcal{L}^2\) integrands on \([0,T]\)?