Having trouble finding expected value and variance of a stochastic process defined by SDE:

$dX_{t} = a X_{t} dt + b dB_{t}$

$X_0 = x$, $a$ and $b$ are constant values, $B_t$~$N(0,t)$

Thank you for any help or pointers. I will now post my progresses:

1) Considering an auxiliary process $Z_t=X_te^{-at}$

2) Using Ito's formula to derive $dZ_t$ omitting quadratic covariance term due to it being between a stochastic and a determinist process (=0):

$dZ_t = e^{-at} dX_t+X_t(-ae^{-at}) dt$

3) Plugging in $dX_t$:




4) Applying the integral on both sides ( change of variable from t to s inside the integral)

$\int_{0}^{t}dZ_s = b \int_{0}^{t}e^{-as}dB_s $

$Z_t-Z_0 = b \int_{0}^{t}e^{-as}dB_s $

5) Plugging back in $Z_t=X_te^{-at}$ and isolating $X_t$


$X_t=X_0e^{at} + be^{at}\int_{0}^{t}e^{-as}dB_s$

6) Applying expectations

$\mathbb{E}(X_t)=xe^{at} + be^{at}\,\mathbb{E}(\int_{0}^{t}e^{-as}dB_s)$

I end up with this expression, now I know that I should use the Doléans exponential to prove that the stochastic integral is 0. But how can I proceed further to recover both expected value and variance?

  • $\begingroup$ So what happens if you plug in $dX_t$? (And I suppose $(B_t)_{t \geq 0}$ is a Brownian motion. If so, the statement "$B_t \sim N(0,1)$" is not correct.) $\endgroup$ – saz Jan 11 '15 at 17:01
  • $\begingroup$ You plugged in $dX_t$, so there should be no "$dX_t$" in the last 3 lines. (Note that the expression $dB_t dX_t$ doesn't even make sense...) Consequently, you end up with $$dZ_t = b dB_t.$$ Solve it! $\endgroup$ – saz Jan 11 '15 at 17:07
  • $\begingroup$ yes, thank you I'll fix the mistakes you pointed out and try to solve it, kinda new to Latex and stoch calculus, please bear with me $\endgroup$ – Clemente Cortile Jan 11 '15 at 17:09
  • $\begingroup$ You are welcome. (There is a typo in my previous comment, it should read $dZ_t = b e^{-a t} \, dB_t$.) $\endgroup$ – saz Jan 11 '15 at 17:24
  • $\begingroup$ So far your calculations are correct. What does this tell you about $X_t$? $\endgroup$ – saz Jan 11 '15 at 17:43

Hint: The stochastic integral

$$M_t := \int_0^t e^{-as} \, dB_s$$

is a martingale. Hence, $\mathbb{E}M_t = \mathbb{E}M_0=0$. In order to calculate the variance of $X_t$ use Itô's isometry.

Alternative approach: Since stochastic integrals are martingales, we have

$$\mathbb{E}X_t- \mathbb{E}X_0 = \int_0^t a \cdot \mathbb{E}X_s \, ds,$$

i.e. $m(t) := \mathbb{E}X_t$ solves the ordinary differential equation (ODE)

$$m'(t) = a \cdot m(t) \qquad m(0) = \mathbb{E}X_0.$$

The (unique) solution is $m(t) =e^{at} \mathbb{E}X_0$. Similarly, using Itô's formla, one can show that

$$\mathbb{E}(X_t^2)-\mathbb{E}(X_0^2) = \int_0^t \left( 2a \mathbb{E}(X_s^2) + b^2 \right) \, ds.$$

Consequently, $\sigma(t) := \mathbb{E}(X_t^2)$ solves

$$\sigma'(t) = 2a \sigma(t)+b, \qquad \sigma(0) = \mathbb{E}(X_0^2).$$

Solving this (linear) ODE yields $\mathbb{E}(X_t^2)$.

Remark: The process $(X_t)_{t \geq 0}$ is called Ornstein-Uhlenbeck process.

  • $\begingroup$ How is the expected value of a process defined? Does it depend on $t$? $\endgroup$ – user415535 Sep 13 '17 at 10:03
  • 1
    $\begingroup$ @user21312 Yes, it does depend on $t$; the expected value is the mapping $m(t) := \mathbb{E}(X_t)$. $\endgroup$ – saz Sep 13 '17 at 12:51

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.