3
$\begingroup$

In this paper by Jaeckel (2006), he derives the asymptotics for the option price $b$ as:

\begin{align*} \lim_{\sigma \to \infty}b= e^{\theta x/2} - \frac{4}{\sigma}\cdot \phi(\sigma/2) \tag{2.7}\\ \lim_{\sigma \to 0} b = \iota +x \cdot \phi(\frac{x}{\sigma}) \cdot \left(\frac{\sigma}{x}\right)^3 \tag{2.8} \end{align*}

For: $$\iota = h(\theta x)\cdot\theta\cdot\left(e^{\frac{x}{2}}-e^{-\frac{x}{2}}\right),$$ where $h(\cdot)$ is the heaviside function.

I get that the Abramovich-Stegun approximation for the gaussian cumulative distribution function(CDF) $\Phi(z)$ is perhaps used to derive expressions (2.7) and (2.8). But, I couldn't follow through and derive these myself.

I would like to ask for some help, to understand how Jaeckel derives this expressions.

$\endgroup$

1 Answer 1

2
$\begingroup$

As you mentioned, we know the Abramovich-Stegun approximation for the CDF $$\Phi(z) = h(z) − \dfrac{\varphi(z)}{z} \left[ 1 - \dfrac{1}{z^2} + \mathcal{O}\left(z^{-4} \right)\right], \quad \text{for} \; |z| \to \infty.$$

Moreover, note that we can develop the term $$ e^{\pm x/2} \varphi \left(\theta \left[\dfrac{x}{\sigma} \pm \dfrac{\sigma}{2} \right] \right) = \dfrac{1}{\sqrt{2\pi}}\exp \left\lbrace \pm \dfrac{x}{2} - \dfrac{1}{2}\left(\dfrac{x^2}{\sigma^2} + \dfrac{\sigma^2}{4}\right) \mp \dfrac{x}{2}\right\rbrace = \sqrt{2\pi} \varphi \left( \dfrac{x}{\sigma}\right) \varphi \left(\dfrac{\sigma}{2} \right), $$ where $ \sqrt{2\pi} \varphi \left( \dfrac{x}{\sigma}\right) \to 1 $ for $\sigma \to \infty$ and $ \sqrt{2\pi} \varphi \left( \dfrac{\sigma}{2}\right) \to 1 $ for $\sigma \to 0$.

I will split the derivation in the two cases, $\sigma \to 0$ and $\sigma \to \infty$.

First case: $\sigma \to 0$

In the $\sigma\to0$ case this means $$ e^{\pm x/2}\Phi \left(\theta \left[\dfrac{x}{\sigma} \pm \dfrac{\sigma}{2} \right]\right) \simeq e^{\pm x/2} \left\lbrace h(\theta \cdot x) - \dfrac{\varphi(\theta \left[\dfrac{x}{\sigma} \pm \dfrac{\sigma}{2} \right])}{\theta \left[\dfrac{x}{\sigma} \pm \dfrac{\sigma}{2} \right]} + \mathcal{O}\right\rbrace. $$

Then, in the expression for $b$, we have $$ \begin{aligned} b & = \theta \left\lbrace e^{x/2}\Phi \left(\theta \left[\dfrac{x}{\sigma} + \dfrac{\sigma}{2} \right]\right) − e^{−x/2} \Phi \left(\theta \left[\dfrac{x}{\sigma} - \dfrac{\sigma}{2} \right]\right)\right\rbrace \\ & \simeq \theta \cdot h(\theta \cdot x) \left( e^{x/2} − e^{−x/2} \right) + \theta \cdot \varphi(x/\sigma) \left[ \dfrac{-\theta}{\dfrac{x}{\sigma} + \dfrac{\sigma}{2}} + \dfrac{\theta}{\dfrac{x}{\sigma} - \dfrac{\sigma}{2}} \right] \end{aligned} $$ and doing simple algebra you can see that the last term in brackets equals (in the $\sigma \to 0$ limit) $$ \left[ \dfrac{-1}{\dfrac{x}{\sigma} + \dfrac{\sigma}{2}} + \dfrac{1}{\dfrac{x}{\sigma} - \dfrac{\sigma}{2}} \right] = \dfrac{x}{\left( x / \sigma \right)^3}. $$

Therefore, we have $$ b \simeq \theta \cdot h(\theta \cdot x) \left( e^{x/2} − e^{−x/2} \right) + \varphi(x/\sigma) \cdot x \cdot \left(\sigma / x\right)^3. $$

Second case: $\sigma \to \infty$

In the $\sigma\to\infty$ case this means $$ e^{\pm x/2}\Phi \left(\theta \left[\dfrac{x}{\sigma} \pm \dfrac{\sigma}{2} \right]\right) \simeq e^{\pm x/2} \left\lbrace h(\pm \theta) - \dfrac{\varphi(\theta \left[\dfrac{x}{\sigma} \pm \dfrac{\sigma}{2} \right])}{\theta \left[\dfrac{x}{\sigma} \pm \dfrac{\sigma}{2} \right]}+ \mathcal{O} \right\rbrace. $$

Then, in the expression for $b$, we have $$ \begin{aligned} b & = \theta \left\lbrace e^{x/2}\Phi \left(\theta \left[\dfrac{x}{\sigma} + \dfrac{\sigma}{2} \right]\right) − e^{−x/2} \Phi \left(\theta \left[\dfrac{x}{\sigma} - \dfrac{\sigma}{2} \right]\right)\right\rbrace \\ & \simeq \theta \left( e^{x/2} h(\theta \cdot x) − h(-\theta \cdot x) e^{−x/2} \right) + \theta \cdot \varphi(\sigma / 2) \left[ \dfrac{-\theta}{\dfrac{x}{\sigma} + \dfrac{\sigma}{2}} + \dfrac{\theta}{\dfrac{x}{\sigma} - \dfrac{\sigma}{2}} \right] \end{aligned} $$

similarly to the previous case, now in the $\sigma \to \infty$ limit, we have $$ \left[ \dfrac{-1}{\dfrac{x}{\sigma} + \dfrac{\sigma}{2}} + \dfrac{1}{\dfrac{x}{\sigma} - \dfrac{\sigma}{2}} \right] = - 4 / \sigma. $$

Therefore, we have $$ b \simeq e^{\theta \cdot x/2} - 4 / \sigma \cdot \varphi(\sigma/2), $$ where I have used the definition of the Heaviside to cancel one of the two terms with $h(\cdot)$.

Hope this helps!

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.