I am struggling to solve the following SDE, $$dX_t = aX_t dt + b \sqrt{X_t} dW_t$$ where $a$ and $b$ are constants, and $X_0$ is given.

I know to solve $dX_t = aX_t dt + b dW_t$ or $dX_t = a dt + b \sqrt{X_t} dW_t$, but when both $X_t$ and $\sqrt{X_t}$ are introduced, how can I solve it?

  • $\begingroup$ This is a square-root diffusion process, and in finance, it is commonly known as the Cox-Ingersoll-Ross model (CIR) model. It cannot be solved exactly. At best, you can give it's moments like the expected value, variance, and distribution, which I believe is a $\chi^2$ distribution. $\endgroup$ – Dr. Ikjyot Singh Kohli Mar 4 '18 at 17:03
  • $\begingroup$ Indeed, the complete question is to find $E(X_t)$ and $Var(X_t)$. But when using $d(e^{at}X_t)$, I don't know how to cancel out $X_t$ or $\sqrt{X_t}$ on right hand side. What is the proper way to do that? $\endgroup$ – SkyDreamer Mar 4 '18 at 17:53
  • $\begingroup$ See answer below. $\endgroup$ – Dr. Ikjyot Singh Kohli Mar 4 '18 at 18:10

What you can do is the following.

Given $dX = a X dt + b \sqrt{X} dW$ (I have left off the subscript $t$ for ease of typing!), given some initial condition $X(0) = X_0$, the solution to this SDE is found by the following:

Multiply both sides $e^{-at}$ and you get:

$e^{-at} dX = a e^{-at} X dt + b e^{-at} \sqrt{X} dW$,

$e^{-at} dX - ae^{-at} Xdt = be^{-at} \sqrt{X} dW$,

$d[e^{-at} X(t)] = be^{-at} \sqrt{X} dW$

Integrating both sides on some interval, say $[0,t]$, we get:

$e^{-at} X(t) - X_0 = b \int_{0}^{t} e^{-as} \sqrt{X(s)} dW(s)$

You can now use this solution to compute the expectation value and variance.


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.