Consider a GBM : $$S(t) = S(0)\exp\left({(\mu-\frac{1}{2}\sigma^2) t + \sigma W_t}\right)$$ $$d\log S(t) = (\mu-\frac{1}{2}\sigma^2) t + \sigma dW_t$$ $$\frac{d S(t)}{S(t)} = \mu t + \sigma dW_t$$ where $W_t$ is the value of random Brownian Motion with difference distributed standard normal.

I have seen that $$\mathbb{E}(S_t)= S_0e^{\mu t}$$ $$\operatorname{Var}(S_t)= S_0^2e^{2\mu t} \left( e^{\sigma^2 t}-1\right)$$ And I know that the median is $$S(0)e^{(\mu-\frac{1}{2}\sigma^2)t}$$, But I'm looking for an expression for the skewness (and kurtosis) if there is one.

I think the probability density function of a ''St'' is: $f_{S_t}(s; \mu, \sigma, t) = \frac{1}{\sqrt{2 \pi}}\, \frac{1}{s \sigma \sqrt{t}}\, \exp \left( -\frac{ \left( \ln s - \ln S_0 - \left( \mu - \frac{1}{2} \sigma^2 \right) t \right)^2}{2\sigma^2 t} \right).$

Many thanks!

  • $\begingroup$ What is the skewness of a Lognormal random variable? $\endgroup$ – wolfies Aug 21 '15 at 7:12

To calculate the skewness (in fact, every order moments of GBM), we need Ito's lemma, https://en.wikipedia.org/wiki/It%C3%B4%27s_lemma, which states that for an Ito diffusion $$ dS_t =\mu_tdt+\sigma_tdW_t $$ and a twice-differentiable scalar function $F(t, S)$, we have $$ \begin{align} F(t,S_t)=F(0, S_0)&+\int ^t _0 F_1 (s, S)ds\\ &+\int ^t _0 F_2 (s, S)\mu_sds\\ &+\int ^t _0 F_2 (s, S)\sigma_sdW_s\\ &+\dfrac {1}{2}\int ^t _0 F_{22} (s, S)\sigma_s ^2ds\\ \end{align} $$ for all $t$. These four integrals come from Taylor expansion and the property of Brownian motion that $(dW_t)^2=dt$.

For geometric Brownian motion, we see that $$ \mu_t =\mu S_t, \sigma_t =\sigma S_t $$ moreover, to calculate skewness, we let $F(t, S)=S^3$, which doesn't depend on $t$. Then Ito's lemma gives $$ S^3_t=S_0^3 + \int_0^t (3\mu +3\sigma^2)S_s^3ds+ \int_0^t 3\sigma S_s^3dW_s $$ taking expectations, we get $$ \mathbb E S^3_t= S_0^3+ \int_0^t (3\mu +3\sigma^2)\mathbb E S_s^3ds $$ because the expectation of stochatic integral is always zero (provided $S_s^3$ is integrable).

Now let $h(t)=\mathbb E S^3_t$, then the above equation is $$ h(t)=S_0^3+(3\mu +3\sigma^2)\int_0^t h(s)ds $$ it's obvious: $$ h(t)=S_0^3 e^{(3\mu +3\sigma^2)}=\mathbb E S^3_t $$

Now we have all gradients for skewness: $\mathbb E S_t, \operatorname{Var} S_t, \mathbb E S^3_t$: $$ \mathbb E [S_t-\mathbb E S_t]^3=\mathbb E S_t^3-3\mathbb E S_t\operatorname{Var} S_t-(\mathbb E S_t)^3 $$ and the result follows: $$ S_0^3 e^{3\mu t}[e^{3\sigma^2 t}-3e^{\sigma^2 t}+2] $$

| cite | improve this answer | |

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.