0
$\begingroup$

Consider the following problem, from Bjork's Arbitrage Theory in Continuous Time:

Consider the standard Black-Scholes model. Derive the arbitrage free price process for the $T$-claim $\mathcal{X}$ where $\mathcal{X}$ is given by $\mathcal{X}=\{S(T)\}^\beta$. Here $\beta$ is a known constant.

My approach.

Let $F(t,s)$ be the price of the claim $\mathcal{X}$ at time $t$, when the underlying spot price is $s$.

The Black-Scholes equation for $F$ is: $$ \begin{align} F_t + rsF_s + \frac12 \sigma^2s^2 F_{ss} - rF &= 0 \\ F(T, S(T)) &= S(T)^\beta. \end{align} $$

It is convenient to make a change of variables of the form $\tilde{F}(t,s) = e^{-rt}F(t,s)$, so that the associated stochastic process is: $$ \begin{align} dX &= rX dt + \sigma X dW \\ X(t) &= s. \end{align} $$

After changing variable to $Y = \log X$ and integrating, I find $$ X(T) = s \exp\left((r-\frac12 \sigma^2)(T-t) + \sigma(W(T) - W(t))\right). $$ So by the Feynman-Kac formula I have: $$ \begin{align} F(t,s) &= e^{-r(T-t)}\mathbb{E}\left[X(T)^\beta\right] \\ &= e^{-r(T-t)}\int_{-\infty}^{\infty}s^\beta e^{\beta z} \exp\left(-\frac12 \frac{(z - (r-\frac12\sigma^2)(T-t))^2}{\sigma^2(T-t)}\right) dz, \end{align} $$ which after some computation gives, if I did not make any mistake: $$ F(t,s)=e^{-r(T-t)}s^\beta \exp\left(\frac12\sigma^2\beta^2(T-t) + (r-\frac12\sigma^2)\beta(T-t)\right). $$

Does it sound right?

Also, regardless of whether the pricing formula is correct, I am not sure if what I found is really the arbitrage free stochastic process for $\mathcal{X}$.

$\endgroup$
1
$\begingroup$

If $X$ is a lognormally distributed variable, $X = e^{\mu + \nu Z}$, so $\ln X$ has mean $\mu$ and variance $\nu^2$, and $Z$ is normally distributed, then $$ E\left[ X^n \right] = e^{n\mu + \frac{1}{2} n^2\nu^2} $$ This solves your question with $X = S_T/S_t$, $n = \beta$, $\mu = (r -\frac{1}{2} \sigma^2) (T-t)$ and $\nu = \sigma \sqrt{T-t}$ in the Black Scholes world.

See also the following wiki page for other properties of lognormal distribution:

https://en.wikipedia.org/wiki/Log-normal_distribution

$\endgroup$
  • $\begingroup$ Ok, so using the expression for $E[X^n]$ as you suggest, and plugging in $X=S_T/s$, given that $s$ is deterministic I would have: $$ F(t,s) = e^{-r(T-t)} s^\beta E[X^\beta], $$ which gives precisely the formula I found (I consider the discounted value at time $t$ of the risk-neutral expectation). $\endgroup$ – J. D. Sep 11 at 10:09
  • $\begingroup$ @J.D. yes, that's correct. It's good/useful that you derived it yourself. $\endgroup$ – ilovevolatility Sep 11 at 10:50

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.