I want to calculate the expected value and the variance of the stock process log-returns in the Local Volatility setting (and the realized/terminal correlation but let us begin in the one-dimentional setting). First some intro and my progress so far.

Consider the standard local volatility SDE: $$ d S_t = \mu(t) S_t\, dt + \sigma(t,S_t) S_t dW_t $$ with deterministic drift $\mu(t)$ and some deterministic local volatility function $\sigma(t,x)$. Using Ito's lemma we can derive the following SDE for the $\ln S_t$: $$ d \ln S_t = \ln S_0 + \left( \mu(t) - \frac{1}{2}\sigma^2(t,S_t)\right) dt + \sigma(t, S_t)dW_t $$ which can be written in the integral form as $$ \ln S_t = \ln S_0 + \int_0^t \left( \mu(u) - \frac{1}{2}\sigma^2(u,S_u)\right) du + \int_0^t \sigma(u,S_u) dW_u $$

I am interested in the log-returns over some (small) time interval $\Delta t$, thus, from the above we get $$ \ln\left(\frac{S_{t+\Delta t}}{S_t}\right) = \ln(S_{t+\Delta t}) - \ln(S_{t}) = \\ = \int_t^{t+\Delta t} \left( \mu(u) - \frac{1}{2}\sigma^2(u,S_u)\right) du + \int_t^{t+\Delta t} \!\! \sigma(u,S_u) dW_u $$

Proceeding to the quantities that I want to calculate, namely, the expected value and variance of the log-returns, we have $$ \mathbb{E}\left[ \ln\left(\frac{S_{t+\Delta t}}{S_t}\right) \right] =\\ = \mathbb{E}\left[ \int_t^{t+\Delta t} \left( \mu(u) - \frac{1}{2}\sigma^2(u,S_u)\right) du + \int_t^{t+\Delta t} \!\! \sigma(u,S_u) dW_u \right] = \\ = \int_t^{t+\Delta t} \mu(u) du - \frac{1}{2} \, \mathbb{E} \left[ \int_t^{t+\Delta t} \sigma^2(u,S_u) du\right] + \mathbb{E} \left[ \int_t^{t+\Delta t} \!\! \sigma(u,S_u) dW_u \right] $$ Since the last inegral is zero, we arrive at the following \begin{eqnarray}\label{ev} \mathbb{E}\left[ \ln\left(\frac{S_{t+\Delta t}}{S_t}\right) \right] = \int_t^{t+\Delta t} \mu(u) du - \frac{1}{2} \, \int_t^{t+\Delta t} \mathbb{E} \left[ \sigma^2(u,S_u) \right] du = \quad ... \,\, ??? \end{eqnarray} Here is the place where I get stuck. Question 1: Is there a way to further simplify the second integral? It comes back in the variance derivation below, where I continue with this form of the expected value.

Further, the variance calculation goes as follows: $$ \textrm{var} \left[ \ln\left(\frac{S_{t+\Delta t}}{S_t}\right) \right] = \mathbb{E}\left[ \ln\left(\frac{S_{t+\Delta t}}{S_t}\right) - \mathbb{E}\left[ \ln\left(\frac{S_{t+\Delta t}}{S_t}\right) \right] \right]^2 = \\ = \mathbb{E}\left[ \int_t^{t+\Delta t} \left( \mu(u) - \frac{1}{2}\sigma^2(u,S_u)\right) du + \int_t^{t+\Delta t} \!\! \sigma(u,S_u) dW_u \\ - \int_t^{t+\Delta t} \mu(u) du + \frac{1}{2} \, \int_t^{t+\Delta t} \mathbb{E} \left[ \sigma^2(u,S_u) \right] du \right]^2 = \\ = \mathbb{E}\left[ \int_t^{t+\Delta t} \!\! \sigma(u,S_u) dW_u -\frac{1}{2} \int_t^{t+\Delta t} \left( \sigma^2(u,S_u) - \mathbb{E} \left[ \sigma^2(u,S_u) \right] \right) du \right]^2 = ... $$ Now, we take the square and get the three components $$ ... = \mathbb{E} \left[ \int_t^{t+\Delta t} \!\! \sigma(u,S_u) dW_u \right]^2 \\ - \mathbb{E} \left[ \int_t^{t+\Delta t} \!\! \sigma(u,S_u) dW_u \times \int_t^{t+\Delta t} \left( \sigma^2(u,S_u) - \mathbb{E} \left[ \sigma^2(u,S_u)\right]\right) du \right] \\ + \frac{1}{4} \, \mathbb{E} \left[ \int_t^{t+\Delta t} \left( \sigma^2(u,S_u) - \mathbb{E} \left[ \sigma^2(u,S_u)\right]\right) du \right]^2 $$

By virtue of the Ito's isometry the first component equals $$ \mathbb{E} \left[ \int_t^{t+\Delta t} \!\! \sigma(u,S_u) dW_u \right]^2 = \int_t^{t+\Delta t} \mathbb{E} \left[ \sigma^2(u,S_u) \right] du $$ which is the integral we have already seen in the expected value formula. Thus, the whole variance of the log-return is (so far) given by $$ \textrm{var} \left[ \ln\left(\frac{S_{t+\Delta t}}{S_t}\right) \right] = \int_t^{t+\Delta t} \mathbb{E} \left[ \sigma^2(u,S_u) \right] du \\ - \mathbb{E} \left[ \int_t^{t+\Delta t} \!\! \sigma(u,S_u) dW_u \times \int_t^{t+\Delta t} \left( \sigma^2(u,S_u) - \mathbb{E} \left[ \sigma^2(u,S_u)\right]\right) du \right] \\ + \frac{1}{4} \, \mathbb{E} \left[ \int_t^{t+\Delta t} \left( \sigma^2(u,S_u) - \mathbb{E} \left[ \sigma^2(u,S_u)\right]\right) du \right]^2 $$

Question 2: In general, how to proceed further? Does the expected value of the product of the integrals simplify to something? Third component resembles the variance of the squared local volatility function - can we use it somehow?

Note that in case of constant volatility the variance of the log-return simplifies to $\sigma^2 \Delta t $, which is exactly what we define as volatility in the Black-Scholes setting. But I want to know what is the (empirical/realized) volatility in case of the Local Volatility model setting (as defined by the time scaled variance of the log-returns of the stock process).

Any advice/hint (also partial) would be appreciated!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.