I'm new to stochastic calculus and I did an exercise but I don't know if it is correct, so I need somebody with more experience to check if it is true.

I am trying to compute the variance of the following random variable:

$$Z=\int _0^T e^{W_t} dW_t$$

So we have:

$\text{Var}(Z)=\text{Var}\left(\int _0^T e^{W_t} dW_t\right)$

By Itō's isometry we have:

$$\mathbb{E}\left[\int _0^T e^{2W_t} dt\right]$$

we can then bring inside the expectation to get:

$$\int _0^T \mathbb{E}\left[e^{2W_t}\right] dt = \int_0^T e^{2t} dt = \frac{e^{2T}}{2}-\frac{1}{2}$$

Moreover, if the above result is correct, what should I get instead of the problem asked me to compute

$$\text{Var}\left(\int _0^T e^{W_t} dt \right)$$

It should simply be the variance of a lognormal distributed random variable, computed in the extrema of the interval, or not?

  • 2
    $\begingroup$ $Var(Z) = E(Z^2) - [E(Z)]^2$. $\endgroup$ – Gordon May 31 '15 at 0:48
  • $\begingroup$ I know this basic stuff, please tell me if what I did is correct. I've just applied the basic relationships of stochastic calculus $\endgroup$ – james42 May 31 '15 at 0:51
  • $\begingroup$ That is correct. $\endgroup$ – Gordon May 31 '15 at 1:25
  • $\begingroup$ @Gordon What about the second question? $\endgroup$ – BCLC Dec 29 '16 at 0:51
  • $\begingroup$ What does "if the above result is correct, what should I get instead of the problem asked me to compute" mean? $\endgroup$ – SRKX Dec 29 '16 at 10:02

To compute the variance $$\text{Var}\left(\int _0^T e^{W_t} dt \right),$$ we need to compute \begin{align*} E\left( \left(\int _0^T e^{W_t} dt \right)^2 \right) &= \int_0^T\!\!\!\!\int_0^T E\left(e^{W_s+W_t} \right) ds\,dt. \end{align*} Note that, for $0 \le s, t \le T$, \begin{align*} W_s+W_t = \begin{cases} W_t -W_s + 2 W_s, & \text{ if } s \le t,\\ W_s -W_t + 2 W_t, & \text{ else}. \end{cases} \end{align*} That is, as a sum of two independent normal random variables, $W_s+W_t$ is normal, with mean $0$ and variance \begin{align*} \text{Var}(W_s+W_t) = \begin{cases} t+3s, & \text{ if } s \le t,\\ s+3t, & \text{ else}. \end{cases} \end{align*} Then \begin{align*} E\left( \left(\int _0^T e^{W_t} dt \right)^2 \right) &= \int_0^T\!\!\!\!\int_0^T E\left(e^{W_s+W_t} \right) ds\,dt\\ &=\int_0^T\left[\int_0^t E\left(e^{W_s+W_t} \right) ds+\int_t^T E\left(e^{W_s+W_t} \right) ds\right]dt\\ &=\int_0^T\left[\int_0^t e^{\frac{1}{2}t + \frac{3}{2}s} ds+\int_t^T e^{\frac{1}{2}s + \frac{3}{2}t} ds\right]dt. \end{align*} The remaining is simple calculus.


$$Var(\int _0^T e^{W_t} dt)$$

$$= E[(\int _0^T e^{W_t} dt)^2] - (E[\int _0^T e^{W_t} dt])^2$$


$$E[\int _0^T e^{W_t} dt] = \int _0^T E[e^{W_t}] dt$$

Recall that $W_t$ is normal. use mgf

As for

$$E[(\int _0^T e^{W_t} dt)^2]$$

I'll try following this:

$$E[(\int_0^T e^{W_t} dt)^2]$$

$$ = E[(\int_0^T e^{W_t} dt)(\int _0^T e^{W_s} ds)]$$

$$ = E[\int_0^T \int_0^T e^{W_t} e^{W_s} dt ds]$$

$$ = \int_0^T \int_0^T E[e^{W_t} e^{W_s}] dt ds$$

Without loss of generality, suppose $s < t$. Then by considering the exponential martingale, we have

$$E[e^{W_t} e^{W_s}] = E[E[e^{W_t} e^{W_s}|\mathscr F_s]]$$

$$= E[e^{W_s}E[e^{W_t} |\mathscr F_s]]$$

$$= E[e^{W_s}e^{\frac{t}{2}}E[e^{\frac{-t}{2}}e^{W_t} |\mathscr F_s]]$$

$$= E[e^{W_s}e^{\frac{t}{2}}e^{\frac{-s}{2}}e^{W_s}]$$

$$= e^{\frac{t}{2}}e^{\frac{-s}{2}}E[e^{W_s}e^{W_s}]$$

$$= e^{\frac{t}{2}}e^{\frac{-s}{2}}E[e^{2W_s}]$$

Note that $2W_s$ is normal too. Use mgf again


$$ \int_0^T \int_0^T E[e^{W_t} e^{W_s}] dt ds$$

$$ = \int_0^T e^{\frac{t}{2}} dt \int_0^T e^{\frac{-s}{2}}E[e^{2W_s}] ds$$

  • 1
    $\begingroup$ $W_t$ is not log-normal. $e^{W_t}$ is. $\endgroup$ – SRKX Dec 29 '16 at 9:55
  • 1
    $\begingroup$ Also, the conclusion is not clear. You should keep the left part of the equation at least in the first line. $\endgroup$ – SRKX Dec 29 '16 at 10:04

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.