1
$\begingroup$

I'm reading Shreve's Stochastic Calculus for Finance II. On page 191, Exercise 4.6, we are given the problem

Exercise 4.6. Let $S(t)=S(0)\exp\Big \{\sigma W(t)+(\alpha-\frac{1}{2}\sigma^2)t\Big\}$ be a geometric Brownian motion. Let $p$ be a positive constant. Compute $\mathrm{d}(\{S(t)\}^p)$, the differential of $S(t)$ raised to the power of $p$.

The details of how I solve this problem aren't too relevant to the question, so a reader can skip to The Question.

Solving the Problem

I can solve this problem with a direct application of the Itô-Doeblin formula (page 138). I reproduce that formula here for completeness.

Theorem 4.4.1 (Itô-Doeblin formula for Brownian motion). Let $f(t, x)$ be a function for which the partial derivatives $f_t(t, x)$, $f_x(t, x)$, and $f_{xx}(t, x)$ are defined and continuous, and let $W(t)$ be a Brownian motion. Then, for every $T\geq0$,

$$f(T, W(T)) = f(0,W(0)) + \int^{T}_0 f_t(t, W(t)) \mathrm{d}t + \int^{T}_0 f_x(t, W(t))\mathrm{d}W(t)+\frac{1}{2}\int^T_0f_{xx}(t, W(t))\mathrm{d}t\text{.}\tag{1}$$

We define

$$f(t, x) = \Big( S(0)e^{\sigma x + (\alpha-\frac{1}{2}\sigma^2)t}\Big)^p\text{.}\tag{2}$$

We then directly apply Theorem 4.4.1 to get \begin{align} \mathrm{d}f(t, W(t)) = \mathrm{d}(\{S(t)\}^p) &= pS(0)^p(\alpha+(\frac{p-1}{2})\sigma^2) e^{p(\sigma W(t)+(\alpha-\frac{1}{2}\sigma^2)t)}\mathrm{d}t + pS(0)^p(\sigma)e^{p(\sigma W(t)+(\alpha-\frac{1}{2}\sigma^2)t)}\mathrm{d}W(t) \\ &= pS(t)^p\Big[\sigma \mathrm{d}W(t) + (\alpha + \frac{p-1}{2})\mathrm{d}t \Big]\tag{3}\end{align}

or, to put it in integral form,

$$\int_0^T (S(t))^p \mathrm{d}t = \int_0^T pS(t)^p\sigma \mathrm{d}W(t) + \int_0^T pS(t)^p(\alpha + \frac{p-1}{2})\mathrm{d}t\text{.}\tag{4}$$

The Question

The math involved in the solution above makes sense, and I believe this solution matches the author's intent. However, I don't see why the above work has been productive. When I "compute" something, I imagine that you execute an algorithm to solve a problem (perhaps just applying the Itô-Doeblin formula was the "computation"), or, in most cases, put a problem in a "closed-form" where known algorithms can then solve it. Often you simplify a formula to remove complicating parts.

I don't see why I am done when I write line $(4)$. I am not sure how you would compute $(4)$ or ever use $(4)$. I am not sure why $(4)$ is a better form than just writing $\mathrm{d}\{S(t)\}^p)$. Assuming you had some interest in $\mathrm{d}\{S(t)\}^p)$, how would you ever use $(4)$? What would be the inputs? What would be the outputs? Would you use it with a computer, or just paper and pencil? Is there any use of this formula that isn't just getting a distribution of $\mathrm{d}\{S(t)\}^p)$ for time $t$ when you have all of the information up to $t-1$? Also, the book suggests integrals w.r.t. $\mathrm{d}t$ are "just Lebesgue" integrals, but when the integrand involves Brownian Motion, I don't see why that is helpful.

$\endgroup$
1
  • 2
    $\begingroup$ If Brownian motion is in the integrand but is not the integrator, then the integral can be interpreted in the pathwise sense. $\endgroup$ Apr 21 at 17:47
2
$\begingroup$

Note that SDE (4) does have a "closed-form" representation.

Let $X$ be $$X := S^p, $$

so (4) is a geometric Brownian motion SDE $$dX = (p\alpha + 2^{-1}p(p-1) \sigma^2) X dt + p \sigma X dW, $$

which, again due to Ito Lemma, is equivalent to

$$ d \ln X = (p\alpha + 2^{-1}p(p-1) \sigma^2 - 2^{-1}p^2 \sigma^2) dt + p \sigma dW $$

or

$$ d \ln X = (p\alpha - 2^{-1} p\sigma^2) dt + p \sigma dW $$

and, finally:

$$ X_t = X_0 \exp (p \sigma W_t + p(\alpha - 2^{-1} \sigma^2)t). $$

Of course, this was expected from the beginning, as $S$ was given in "closed-form":

$$ \left( S_0 \exp (\sigma W_t + (\alpha - 2^{-1} \sigma^2)t) \right)^p = S_0^p \exp (p \sigma W_t + p(\alpha - 2^{-1} \sigma^2)t)$$

$\endgroup$

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.