Using simple delta-probability decomposition, the price European call options a non- dividend paying asset can be computed as

\begin{equation} C(T,K) = {S_0}{\rm{ }}{\Pi _1} - {e^{ - rT}}K{\rm{ }}{\Pi _2}, \end{equation}


\begin{equation} {\Pi _1} = \frac{1}{2} + \frac{1}{\pi }\int_0^\infty {{\mathop{\rm Re}\nolimits} \left[ {\frac{{{e^{ - iw\ln (K)}}{\psi _{\ln {S_T}}}(w - i)}}{{iw{\psi _{\ln {S_T}}}( - i)}}} \right]} \;dw, \end{equation}

\begin{equation}\label{pi2} {\Pi _2} = \frac{1}{2} + \frac{1}{\pi }\int_0^\infty {{\mathop{\rm Re}\nolimits} \left[ {\frac{{{e^{ - iw\ln (K)}}{\psi _{\ln {S_T}}}(w)}}{{iw}}} \right]} \;dw, \end{equation}

and where ${\psi _{\ln {S_T}}}$ is the characteristic function of the log-asset price. For instance, for the Heston and Variance Gamma models, the corresponding $\psi^{H}_{\ln {S_T}}$ and $\psi^{VG}_{\ln {S_T}}$ are given by:

  • Variance Gamma

\begin{equation} \psi _{\ln ({S_t})}^{VG}(w) = {\left( {\frac{1}{{1 - i\theta vw + ({\sigma ^2}v/2){w^2}}}} \right)^{t/v}} \end{equation}

  • Heston

\begin{equation} \psi_{\ln(S_t)}^{H} (w) = e^{ C(t,w) \overline{V}+ D(t,w) V_0 +iw \ln(S_0 e^{rt})}, \end{equation} where \begin{eqnarray*} C(t,w) &=& a \left[ r_{-} \cdot t - \frac{2}{\eta^2} \ln \left( \frac{1-g e^{-ht}}{1-g} \right) \right], \\ D(t,w) &=& r_{-} \frac{1-e^{-ht}}{1-g e^{-ht}}, \\ \alpha &=& - \frac{w^2}{2}- \frac{iw}{2}, \quad \beta = a - \rho \eta i w , \quad \gamma = \frac{\eta^2}{^2}, \\ r_{\pm} &=& \frac{ \beta \pm h}{\eta^2}, \quad h= \sqrt{ \beta^2- 4 \alpha \gamma}, \quad g= \frac{r_{-}}{r_{+}}. \end{eqnarray*}

These equations, however, are intended for options on spot prices. If I were interested in options on futures, which modifications should be made to the formulas above?(because it is not as straightforward as simply using $F_0$ instead of $S_0$ and removing the effects of $r$, right?)

  • 2
    $\begingroup$ Why not, future price is a martingale under $\Bbb{Q}$ and has the same instantaneous volatility as the spot price process (pure diffusion assumed). See also: quant.stackexchange.com/questions/24601/… $\endgroup$
    – Quantuple
    Dec 16 '16 at 13:12
  • $\begingroup$ @Quantuple, I say that it cannot be so simple because the expected option payoff should be still discounted, so the $r$ still needs to appear somewhere. $\endgroup$
    – sets
    Dec 16 '16 at 13:31
  • $\begingroup$ Also, I have used the Heston and VG models as examples, but your answer seems to imply that the modifications to price options on futures could be slightly different if I were to use, for instance, the Bates model? $\endgroup$
    – sets
    Dec 16 '16 at 13:39
  • $\begingroup$ No this not what i meant, apologies for my very confusing parentheses. See @LocalVolatility's answer. $\endgroup$
    – Quantuple
    Dec 16 '16 at 16:07

I generally agree with Quantuple's comment.

I explicitly discuss the case of the variance gamma model, though most of this also applies to Heston. First note that in case of the variance gamma model, the characteristic function that you presented is not the one of the logarithmic stock price under the risk-neutral probability measure. Let $X$ be a variance gamma process as in Madan et al. (1998). The characteristic function of $X_t$ is

\begin{equation} \phi_{X_t}(\omega) = \left( 1 - \mathrm{i} \theta \nu \omega + \frac{1}{2} \sigma^2 \nu \omega^2 \right)^{-t / v}; \end{equation}

see Equation (7) in the original paper (note that there was originally a typo in your question). We now look for a drift term $\gamma$, such that the process

\begin{equation} Y_t = \exp \left\{ \gamma t + X_t \right\} \end{equation}

is a martingale. You find that

\begin{equation} \gamma = -\ln \left( \phi_{X_t}(-\mathrm{i}) \right). \end{equation}

Your model for the stock and forward prices respectively is then

\begin{eqnarray} S_t & = & S_0 \exp \left\{ (r + \gamma) t + X_t \right\}\\ F_t(T) & = & F_0(T) \exp \left\{ \gamma t + X_t \right\} \end{eqnarray}

with characteristic functions

\begin{eqnarray} \phi_{\ln \left( S_t \right)}(\omega) & = & \exp \left\{ \mathrm{i} \left( \ln \left( S_0 \right) + r + \gamma \right) \omega t \right\} \phi_{X_t}(\omega)\\ \phi_{\ln \left( F_t(T) \right)}(\omega) & = & \exp \left\{ \mathrm{i} \left( \ln \left( F_0(T) \right) + \gamma \right) \omega t \right\} \phi_{X_t}(\omega) \end{eqnarray}

You can now re-use your general expressions for $\Pi_1$ and $\Pi_2$ in terms terms of the appropriate characteristic function as these are just the corresponding exercise probabilities under the respective measure. Your pricing formula then becomes

\begin{equation} C_0 = e^{-r T} \left( F_0(T) \Pi_1 - K \Pi_2 \right). \end{equation}

If you want to convince yourself that this is correct, then you could work through the detailed steps given e.g. in Schmelzle (2010).


Madan, Dilip B, Peter P. Carr and Eric C. Chang (1998) "The Variance Gamma Process and Option Pricing", European Finance Review, Vol. 2, pp. 79-105

Schmelzle, Martin (2010) "Option Pricing Formulae Using Fourier Transform: Theory and Application", Technical Report


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.