I have carefully reconstructed all the computations that lead to the Heston option pricing formula for a call. I end up with this formula for the "adjusted" probabilities

$$ P_j\left(x,v,T;\ln K\right) = \int_{-\infty}^{\infty}\frac{d\phi}{2\,\pi}\,\frac{e^{-i\,\phi\,\ln\left(K\right)}\,f_j\left(x,v,T;\phi\right)}{i\,\phi}\,d\phi $$

while in the original paper and all textbooks the final formula is expressed as

$$ P_j\left(x,v,T;\ln K\right) = \frac{1}{2}+\int_{0}^{\infty}\frac{d\phi}{\pi}\,\text{Re}\left(\frac{e^{-i\,\phi\,\ln\left(K\right)}\,f_j\left(x,v,T;\phi\right)}{i\,\phi}\,\right)d\phi. $$

I understand that, if $f_j\left(x,v,T;\phi\right)$ is such that

$$ f_j\left(x,v,T;-\phi\right) = \bar{f_j}\left(x,v,T;\phi\right)\quad(1) $$

(where a bar indicates the complex conjugate) then we have immediately that

$$ \int_{-\infty}^{\infty}\frac{d\phi}{2\,\pi}\,\frac{e^{-i\,\phi\,\ln\left(K\right)}\,f_j\left(x,v,T;\phi\right)}{i\,\phi}\,d\phi = \int_0^\infty\frac{d\phi}{\pi}\,\text{Re}\left(\frac{e^{-i\,\phi\,\ln\left(K\right)}\,f_j\left(x,v,T;\phi\right)}{i\,\phi}\,\right)d\phi. $$

My problem is thus twofold. First I cannot see from the definition of $f_j\left(x,v,T;\phi\right)$ that the property (1) is satisfied (neither I have found any kind of discussion in other textbooks) and, second, even if (1) is proved I miss the $\frac{1}{2}$ that appears in the Heston formula.

Any help would be greatly appreciated.

  • $\begingroup$ have you looked at my book "more mathematical finance" chapter 17? $\endgroup$ – Mark Joshi Mar 15 '16 at 22:55

I do not know what are the exact steps you followed, but here's my 2 cents:

\begin{align*} C(S_t,t) &= P(t,T) E^Q [ (S_T - K)^+ ] \\ & = P(t,T) E^Q [ S_T 1_{S_T \geq K} ] - K P(t,T) E^Q [ 1_{S_T \geq K} ] \end{align*}

Which can be re-written (using a change of numéraire)

\begin{align*} C(S_t,t) &= S_t E^{Q^S} [ 1_{S_T \geq K} ] - K P(t,T) E^Q [ 1_{S_T \geq K} ] \\ &= S_t P_1 - K P(t,T) P_2 \end{align*}

where $P_1 = Q^S( S_T \geq K)$ (resp. $P_2 = Q(S_T \geq K)$) are directly related to the cumulative distribution function of the stock price $S_T$ under the measure $Q^S$ (resp. $Q$) where the stock price (resp. the riskless money account) is chosen as numéraire.

Because $P_1$ and $P_2$ share very similar interpretations (up to a change of measure), I simply focus on $P_2$ in what follows.

If we knew the probability density function $q(s_t,v_t,t;S_T=s)$ in closed-form, we could trivially infer the desired cdf by integrating it and hence deduce $P_2$. This is not the case here.

However, we know the characteristic function $f_2$ of the log-stock price $s_T = \ln(S_T)$ under $Q$ (remember that Heston's model is an affine diffusion framework), which by definition writes $f_2(s_t,v_t,t;\phi) = E^Q[e^{i\phi s_T}]$. From this definition comes the equality you mention in your original post.

At this point, there exists a general result, known as the Gil-Pelaez inversion theorem, which tells you how to 'invert' $f_2(\phi)$ (which is basically a Fourier transform) to obtain the desired CDF. Here, the theorem writes

$$Q(s_T < \ln(K) ) = \frac{1}{2} - \frac{1}{\pi} \int_0^\infty \Re \left[ \frac{e^{-i \phi \ln(K)}f_2(\phi))}{i\phi} \right] d\phi$$

Making the missing 1/2 term appear. Noting that $P_2 = 1 - Q(s_T < \ln(K) )$ concludes the demonstration. Similar developments can be applied to obtain $P_1$. You can probably find demonstrations of the Gil-Pelaez inversion theorem in the literature.


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.