Quantitative Finance Stack Exchange is a question and answer site for finance professionals and academics. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I wish to understand some basic fact about the (primitive) simulation of stock prices with geometric Brownian motion.

If $S(t)$ is the stock price at time $t$, and the stock price follows geometric Brownian motion distribution, then it should satisfy $$dS(t) = S(t)\left(\mu dt + \sigma d B(t)\right)$$ where $B(t)$ is a standard linear Brownian motion, and $\mu$ is sometimes called drift. Solving this for $S(t)$ gives $$S(t) = S(0)\cdot \exp\left(\sigma B(t) + \left(\mu-\frac{\sigma^2}{2}\right)t\right)$$

Assuming there is no drift (there is no trend, perhaps), we obtain the following:$$S(t) = S(0)\cdot \exp\left(\sigma B(t) -\frac{\sigma^2}{2}t\right)$$ hence $$\ln\frac{S(1)}{S(0)} \sim \mathcal{N}(0,\sigma^2)-\frac{\sigma^2}{2}$$ and in particular \begin{eqnarray*} \mathbb{E}\left(\ln\frac{S(1)}{S(0)}\right) &=& \int_{-\infty}^\infty x\cdot\mathbb{P}\left(\mathcal{N}(0,\sigma^2) = x +\frac{\sigma^2}{2}\right) dx\\ &=& \int_{-\infty}^\infty x \cdot \frac{1}{\sigma\sqrt{2\pi}}\cdot\exp\left(-\frac{\left(x+\frac{\sigma^2}{2}\right)^2}{2\sigma^2}\right) dx \end{eqnarray*}

It looks like $$\mathbb{E}\left(\ln\frac{S(1)}{S(0)}\right) = -\frac{\sigma^2}{2}$$ which is what I would have expected. However, the more interesting expectation I couldn't compute: \begin{eqnarray*} \mathbb{E}\left(\frac{S_1}{S_0}\right) &=& \int_{-\infty}^\infty x\cdot \mathbb{P}\left(\exp\left(\mathcal{N}(0,\sigma^2)-\frac{\sigma^2}{2}\right)=x\right)dx\\ &=& \int_{-\infty}^\infty x\cdot\mathbb{P}\left(\mathcal{N}(0,\sigma^2)=\ln{x}+\frac{\sigma^2}{2}\right)dx\\ &=& \int_{-\infty}^\infty \frac{x}{\sigma\sqrt{2\pi}}\cdot \exp\left(-\frac{\left(\ln{x} + \frac{\sigma^2}{2}\right)^2}{2\sigma^2}\right)dx = ? \end{eqnarray*}

I would expect that expectation to be 1, as there is no drift. However, I can't do the integration. Is that true to assume that the result of the integral is 1?

share|improve this question
Mathematica says that it evaluates to $0$. – vonjd Jun 2 '14 at 15:04
Thanks, that makes me happy... – Bach Jun 2 '14 at 15:16
Didn't you say that you 'wanted' a $1$? – vonjd Jun 2 '14 at 15:17
Oh, you right, I was thinking about the logarithm... so that actually makes no sense. I might have a calculation error somewhere. – Bach Jun 2 '14 at 15:23
up vote 3 down vote accepted

To solve the expectation directly, you need to remember that a density function is not the same as the probability of the event.

We have, $\frac{S_1}{S_0} \sim \ln \mathcal{N} \left(-\frac{\sigma^2}{2},\sigma\right)$, therefore,

\begin{eqnarray} \mathbb{E}\left(\frac{S_1}{S_0}\right) &=& \int_{-\infty}^\infty x\, f_{\frac{S_1}{S_0}}(x)dx\\ &=&\int_{-\infty}^\infty x \, \frac{1}{x\sqrt{2\pi}\sigma}\exp\left[-\frac{\left(\ln x+\frac{\sigma^2}{2}\right)^2}{2\sigma^2}\right]dx. \end{eqnarray} The usual substitution method aims to manipulate the expression into looking like a standard normal. To do this we use $z=\frac{\ln x + \frac{\sigma^2}{2}}{\sigma}-\sigma$ and we get the substitutions $dx=x \sigma dz$ and $x=\exp(\sigma z + \frac{\sigma^2}{2})$.

\begin{eqnarray} \mathbb{E}\left(\frac{S_1}{S_0}\right) &=& \int_{-\infty}^\infty \frac{1}{\sqrt{2\pi} \sigma} \sigma \exp\left(\sigma z + \frac{\sigma^2}{2}\right) \exp\left( \frac{-z^2 - 2\sigma z -\sigma^2}{2} \right)dz\\ &=&\int_{-\infty}^\infty \frac{1}{\sqrt{2\pi}}\exp\left(-\frac{z^2}{2}\right) dz\\ &=&1. \end{eqnarray}

share|improve this answer
This is great, thanks! – Bach Jun 4 '14 at 12:50

If $S_t = S_0 e^{(\mu-\sigma^2/2)t + \sigma W_t}$, we can compute $$\mathbb{E}^Q\left[S_T\middle\vert \mathcal{F}_0\right] = S_0 e^{r T} = \text{forward price of } S_T \text { at time } 0. $$ To show the details, $\mathbb{E}^Q\left[S_T\middle\vert \mathcal{F}_0\right] = S_0 e^{(r-\sigma^2/2) T} \mathbb{E}^Q\left[e^{\sigma W_T} \middle\vert \mathcal{F}_0\right]$. To compute $\mathbb{E}^Q\left[e^{\sigma W_T} \middle\vert \mathcal{F}_0 \right]$, note that, for $X \sim \mathcal{N}(\mu,\sigma^2)$, $\mathbb{E}\left[e^X \right]=e^{\mu+\sigma^2/2}$.

In your case, since $r=0$ and $T = 1$, $$\mathbb{E}^Q\left[\frac{S_1}{S_0}\middle\vert \mathcal{F}_0\right] = 1. $$

share|improve this answer
Thanks! Could you explain what the superscript $Q$ means? And: when you say, "we can compute", what exactly do you have in mind? – Bach Jun 2 '14 at 19:32
The superscript $Q$ indicates that the expectation is taken over the risk-neutral probability measure [en.wikipedia.org/wiki/Risk-neutral_measure]. To compute $\mathbb{E}^Q\left[ e^{\sigma W_t} \right]$, note that, for $X \sim \mathcal{N}(\mu, \sigma^2)$, $\mathbb{E}\left[ e^X\right] = e^{\mu+\sigma^2/2}.$ – wsw Jun 2 '14 at 19:41
The link is broken. – Bach Jun 2 '14 at 19:42
@Bach: Try this one: en.wikipedia.org/wiki/Risk-neutral_measure – vonjd Jun 3 '14 at 13:48
@Bill, you should copy paste carefully from files (: you included the bracket in the domain link) – user7056 Jun 6 '14 at 21:52

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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