We want to solve:

$dX_t = \alpha X_t dW_t + \sigma X_t dt$

where the initial condition $X_0$ is given and $\alpha$, $\sigma$ are constant.

The solution goes as follows - We rewrite the equation as:

$\frac{dX_t}{X_t} = \alpha dW_t + \sigma dt$

and we can get

$\int_0^t\frac{dX_u}{X_u} = \int_0^t\alpha dW_u + \int_0^t\sigma t = \alpha W_t + \sigma t $

We use Itos formula and set $f(t,x) = f(x) = \ln x$. Why do we do this?

But then we get:

$d \ln(X_t) = \frac{1}{X_t}dX_t - \frac{1}{2} \frac{1}{X_t^2}(dX_t)^2 = \frac{1}{X_t}dX_t - \frac{\alpha^2}{2}dt$

How do get the last step i.e how is:

$\frac{1}{2} \frac{1}{X_t^2}(dX_t)^2 = \frac{\alpha^2}{2}dt$

By this the solution conclude that we can write

$\ln(X_t)-\ln(X_0) = \int_0^t \frac{dX_u}{X_u} - \frac{\alpha^2t}{2}$

Why is this obvious?

The rest of the solution is rather trivial so i let it be out. Thanks!


We use Itô's formula and set $f(t,x) = f(x) = \ln x$. Why do we do this?

Well, if we consider instead of the SDE

$$dX_t = \alpha X_t \, dW_t + \sigma X_t \, dt \tag{1}$$

the "associated" ordinary differential equation

$$dx_t = \sigma x_t \, dt$$

(i.e. we just drop the stochastic integral), then it is well-known that the substitution $y_t := \log x_t$ leads to a very simple ODE which can be solved explicitly. So, if we want to solve the SDE (1), we simply give it a try and use the same substitution as in the deterministic case.

[....] how is $\frac{1}{2} \frac{1}{X_t^2} \, (dX_t)^2 = \frac{\alpha^2}{2} \, dt$.

This follows from the heuristical (!) rules

$$(dW_t)^2 = dt \qquad dt \, dW_t = 0 \qquad (dt)^2 =0$$

and $(1)$; indeed

$$(dX_t)^2 = (\alpha X_t \, dW_t + \sigma X_t \, dt)^2 = \alpha^2 X_t^2 \, dt.$$

By this the solution concludes that we can write $$\ln X_t - \ln X_0 = \int_0^t \frac{dX_u}{X_u} - \frac{\alpha^2}{2} t. \tag{3}$$

We know from the above considerations that

$$d \ln (X_t) = \frac{1}{X_t} \,d X_t - \frac{\alpha^2}{2} \, dt.$$

By the very definition (!) this differential equation is to be understood as an integral equation, i.e.

$$\ln X_t - \ln X_ 0 = \int_0^t \, d\ln (X_u) = \int_0^t \frac{1}{X_u} \, dX_u - \int_0^t \frac{\alpha^2}{2} \, du.$$


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.