First of all, I am a mathematician, so I apologize for my ignorance regarding stochastic calculus. What exactly does an expression like:

$$ \mathbb{E}[dX_tdY_t] $$ here $X_t,Y_t$ are stochastic processes? The differentials $dX_t, dY_t$ are not exactly well defined, so I was wondering what this expression means.

  • $\begingroup$ Though a mathematician does not necessarily know stochastic calculus, but people will usually think you do. $\endgroup$
    – Gordon
    Jul 10, 2019 at 16:25
  • $\begingroup$ @Gordon, to clarify, I understand the meaning of the expression $dX_t = b(t,X_t)dt + \sigma (t,X_t)\mathrm{d}W_t$, but rather, did not know what $\mathbb{E}[dX_t]$ represented. $\endgroup$ Jul 10, 2019 at 16:29

1 Answer 1


In stochastic calculus, expressions of the type: $$ dX_t = a(t, X_t)dt + b(t, X_t) dW_t $$

are called stochastic differential equations.

What the one above means for example is that $X_t$ has the following expression: $$ X_t = X_0 + \int_0^t a(u, X_u)du + \int_0^t b(u, X_u) dW_u $$

The first integral is a regular one, and the second is called a stochastic integral or Ito integral. You can find a rigorous definition of stochastic integrals in any stochastic calculus notebook. It is defined as the limit of a sum over some subdivision when its mesh goes to zero (similar to how the Riemann integral is defined).

See for example:https://en.wikipedia.org/wiki/It%C3%B4_calculus

As for the symbol's definition, I think $\mathbb{E} \left[dX_tdY_t \right]$ denotes the covariation of $X$ and $Y$, which is sometimes denoted $d\langle X, Y\rangle_t$ or $d[X, T]_t$.

This quantity also has a rigorous mathematical definition (also as a sum, resembling that of a covariance, over a partition when the mesh goes to zero). You can think of it as the instantaneous covariance between $X$ and $Y$.

See for example: https://en.wikipedia.org/wiki/Quadratic_variation

  • 1
    $\begingroup$ Thank you, I did not know that this was an expression for the quadratic covariation, I had only seen the brackets before! This clears it up. $\endgroup$ Jul 10, 2019 at 16:30
  • 1
    $\begingroup$ Quick Question: is the term $d\langle X,Y \rangle_t$ supposed to be interpreted as a quadratic covariation, or the time derivative, $\frac{d\langle X , Y\rangle_t}{dt}$ ? $\endgroup$ Jul 11, 2019 at 19:25
  • $\begingroup$ If $X_t = a(t, X_t)dt + b(t, X_t) dW^X_t $ and $Y_t = c(t, Y_t) dt + d(t, Y_t) dW^Y_t$ are two Ito processes, then $d\langle X, Y \rangle_t = \rho_{(W^X, W^Y)}( b(t, X_t)d(t, Y_t) dt$, or equivalently: $\langle X, Y \rangle_t = \int_0^t \rho_{(W^X, W^Y)}( b(u, X_u)d(u, Y_u) du$ I hope this clarifies things for you. $\endgroup$
    – byouness
    Jul 11, 2019 at 20:06
  • 1
    $\begingroup$ yes, it does! Thanks again. $\endgroup$ Jul 11, 2019 at 20:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.