Let $f$ be a continuous function on $[0,\infty)$ and $B_t$ be a standard Brownian motion. Define

$X_t=\int_0^t f(s) dB(s).$

a) Show that $X_t$ is Gaussian and computer its covariance $C(X_s, X_t)$ b) prove that $e^{(X_t - 1/2 C(X_t, X_t))}$ is a continuous time martingale.

I think I know how to show $X_t$ is Gaussian (basically expressing it as a sum of Brownian motion increments each of which is gaussian) is that right? I don't know how to derive the covariance and show that $e^{(X_t - 1/2C(X_t, X_t))}$ is a martingale, any hint would be appreciated.

  • $\begingroup$ "Since there is no assumption on f I think that it could be a random function" There is an "assumption on f", which is that f is deterministic and continuous. $\endgroup$ – Did Oct 19 '14 at 17:55
  • $\begingroup$ So if f is continuous on $[0, \infty)$ it means that it is deterministic ? $\endgroup$ – nick Oct 19 '14 at 17:57
  • $\begingroup$ Oh I see, you are right! but still I don't know how to compute the covariance. $\endgroup$ – nick Oct 19 '14 at 17:58
  • $\begingroup$ Cannot you compute $$E\left(\int f(t)dB_t\int g(t)dB_t\right),$$ for every $(f,g)$? This must be in your notes... $\endgroup$ – Did Oct 19 '14 at 18:01
  • $\begingroup$ This I think is the ito isometry so the answer would be $E(\int_0^tf(t)g(s)ds)$ right? $\endgroup$ – nick Oct 19 '14 at 18:10
  1. Covariance: As @Did pointed out, Itô's formula combined with the polarization identities (i.e. $a \cdot b = \frac{1}{4} ((a+b)^2-(a-b)^2)$) yields $$\mathbb{E} \left( \int g(r) \, dB_r \int h(r) \, dB_r \right) = \mathbb{E} \left( \int g(r) h(r) \, dr \right)$$ whenenever the expression is well-defined. Setting $g(r) := 1_{[0,t]}(r) f(r), h(r) := 1_{[0,s]}(r) f(r)$, we find $$\mathbb{E}(X_t X_s) = \mathbb{E} \left( \int_0^{s \wedge t} f(r)^2 \, dr \right) = \int_0^{s \wedge t} f(r)^2 \, dr$$ where the last inequality holds due to the fact that $f$ is deterministic. Consequently, $$C(s,t) = \int_0^{s \wedge t} f(r)^2 \, dr.$$
  2. Martingale property: This is a straightforward application of Itô's formula. In fact, we have for $F(t,x) := e^{x-G(t)}$, $$\begin{align*}& F(t,X_t) - F(0,0) \\ &\quad= \int_0^t F(s,X_s) \, dX_s - \int_0^t G'(s) F(s,X_s) \, ds + \frac{1}{2} \int_0^t F(s,X_s) f(s)^2 \, ds \end{align*}$$ for any differentiable (deterministic) function $G$. In particular, if we choose $G(t) = \frac{1}{2} C(t,t)$, then this shows $$F(t,X_t)-1 = \int_0^t F(s,X_s) \, dX_s.$$ The right-hand side is a martingale as a stochastic integral with respect to a martingale. This finishes the proof.

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.