1
$\begingroup$

In book The concepts and practice of mathematical finance, in the context of illustrating the stochastic volatility model, the Fourier transform $\hat{P}(\xi, V, T)$ of a European put $P(x, V, T)$ is listed as

$$ \hat{P}(\xi, V, T) = - \frac{K^{i \xi + 1}}{\xi^2 - i \xi} $$

where $x = \log S$ is the logarithm of real world underlying stock values $S$, $\xi$ the respective Fourier variable, $V$ is square vol, $T$ the time maturity, $i$ the imaginary unit. My development of the Fourier integral, using $K - e^x$ as payout for the put, leads to a different result:

$$ \hat{P}(\xi, V, T) = \int e^{i \xi x} ( K - e^x ) dx =\\ \frac{K}{i \xi} e^{i \xi x} - \frac{1}{1 + i \xi} e^{(1 + i \xi) x} = e^{i \xi x} \frac{K + i \xi (K - e^x)}{i \xi - \xi^2} $$

Where have I gone wrong? How to reach the book result?

$\endgroup$
4
$\begingroup$

The generalised Fourier transform $\hat{P}(z)$ of the payoff of a put option with $P(x)=\max\{e^k-e^x,0\}$ is \begin{align*} \hat{P}(z) &= \int_{-\infty}^\infty e^{izx} \left( e^k - e^x \right)^+ \mathrm{d}x \\ &= \int_{-\infty}^k \left( e^ke^{izx}-e^{i(z-i)x} \right)\mathrm{d}x \\ &= \left[ e^k\frac{e^{izx}}{iz} -\frac{e^{i(z-i)x}}{i(z-i)} \right]_{-\infty}^k \\ &= \left( e^k\frac{e^{izk}}{iz} - \frac{e^{i(z-i)k}}{i(z-i)} \right)-0\\ &= \frac{e^{i(z-i)k}}{iz} - \frac{e^{i(z-i)k}}{i(z-i)} \\ &= -\frac{e^{ik(z-i)}}{z(z-i)}. \end{align*} The computation above is only valid if the summand for $x=-\infty$ indeed equals zero. In general, if $z\in\mathbb{R}$, $\lim\limits_{x\to-\infty}e^{ixz}$ does not make sense since $e^{izx}$ merely describes points on the unit circle around the origin. However, if $z=a+ib$ is complex, $e^{izx}=e^{-bx} e^{iax}$ which at least converges to zero as $x\to-\infty$ if $b=\text{Im}(z)<0$ since it contracts the unit circle into the origin. Equivalently, $\lim\limits_{x\to-\infty} |e^{izx}|=\lim\limits_{x\to-\infty} e^{-bx}=0$ if $b=\text{Im}(z)<0$.

Thus, we require $\text{Im}(z)<0$ for the first summand and $\text{Im}(z-i)<0$ for the latter. Both conditions together lead to $\text{Im}(z)<0$. Consequently, the generalised Fourier transform $\hat{P}(z)$ is only well-defined in the open strip $\mathcal{S}_P=\{z\in\mathbb{C}:\text{Im}(z)<0\}$.

The strip of regularity for a call option is $\mathcal{S}_C=\{z\in\mathbb{C}:\text{Im}(z)>1\}$.

Note: it is no coincidence to have $i$ and $0$ as poles of the payoff transform. You can use inversion theorems to see how they relate to $N(d_1)$ and $N(d_2)$ (or more general exercise probabilities).

$\endgroup$
4
  • $\begingroup$ Thank you for specifying the range where the integral does not diverge to $\infty$. So substitute $K = e^{k}$ to obtain the book result. $\endgroup$
    – Giogre
    Jul 18 at 16:30
  • $\begingroup$ @Giogre Precisely. You only need to be careful with the integral boundaries. The calculation itself is easy. And you’re right. I wrote $k=\ln(K)$ for the log-strike price. $\endgroup$
    – Kevin
    Jul 18 at 16:31
  • 1
    $\begingroup$ Nice and concise explanation and description of the integrability condition. $\endgroup$ Jul 18 at 17:38
  • $\begingroup$ @Kermittfrog thank you! For my MSc thesis, I had to compute these payoff transforms for gap options, contingent premium options, power options and other exotic payoffs, haha $\endgroup$
    – Kevin
    Jul 18 at 17:44

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.