Expected value of Ornstein-Uhlenbeck process

231 Views Asked by At

In the paper "The Impact of Jumps in Volatility and Returns" by Nicholas Polson, Bjorn Eraker, and Michael Johannes (2003), the authors state in footnote 6 on page 1273 that, given an Ornstein-Uhlenbeck process of the form: $$dv_t = k(\theta - v_t) dt + \sigma \sqrt{v_t} dW_t$$ where $v_t$ is the process, k is the rate of mean reversion, $\theta$ is the long-run mean, and $\sigma$ is the volatility, we have that:

$$E[v_t] = E[v_0] + E \left[ \int_0^t k(\theta - v_u) du \right] + E \left[ \int_0^t \sigma \sqrt{v_u} dW_u \right]$$

from which they conclude that $$\mu_v = \mu_v + k(\theta - \mu_v)t$$ where $μ_v$ is the expected value of $v$.

I am confused by this conclusion. I remember that the expected value of an Ornstein-Uhlenbeck process is of the form

$$E[v_t] = v_0 e^{-kt} + \theta (1 - e^{-kt})$$ Even if we consider $v_0$ unknown, I do not see how the authors can conclude that $E[v_t]=\mu_v=E[v_0]$.

One possibility is that this is only true for $t$ close to $0$. In this case, the term $\mathbb{E} \left[ \int_0^t k(\theta - v_u) du \right]$ will be small, and the expected value of $v_t$ will be approximately equal to the initial value $v_0$. However, this does not seem to be what the authors are claiming.

Another possibility is that the authors have made a mistake. This is certainly possible, as the footnote is only a few lines long and is not very well-explained.

I would appreciate it if someone could clarify this for me.

1

There are 1 best solutions below

6
On

The access to this paper is restricted so I cannot check if you quote these authors correctly. There is however no reason to assume that the Journal of Finance has printed that nonsensical equation $\mu_v = \mu_v + k(\theta - \mu_v)t\,.$ It should be $$ d\mu_v = k(\theta - \mu_v)\,dt $$ which has the solution that you remember: $$ \mu_v(t)=\mu_v(0)e^{-kt}+\theta(1-e^{-kt})\,. $$