17
$\begingroup$

I want to create a lognormal distribution of future stock prices. Using a monte carlo simulation I came up with the standard deviation as being $\sqrt{(days/252)}$ $*volatility*mean*$ $\log(mean)$. Is this correct?

$\endgroup$
  • 1
    $\begingroup$ Thanks for the help, I figured it was a simple mistake. Actual formula for stdev is sqrt(days/252)*volatility*mean. The mistake I was making is days doesn't include first day. So if you have 5 days of random walk, days=4 in above formula. $\endgroup$ – CptanPanic Feb 3 '11 at 1:34
6
$\begingroup$

I'm not sure I understand, but if you want to compute the variance of $exp(X)$, where $X$ is normally distributed with mean $\mu$ and variance $\sigma^2$, that variance is (from Wikipedia): $$\left(\exp{(\sigma^2)} - 1\right) \exp{(2\mu + \sigma^2)}$$

$\endgroup$
  • $\begingroup$ I don't think I asked my question correctly. Assuming a random walk, starting at $\mu$ and yearly volatility of $\sigma$ . What is $\sigma$ for the distribution (expected prices) in like 1 month? I must be overthinking this, but it doesn't seem like $\sigma / 12 works either. $\endgroup$ – CptanPanic Feb 1 '11 at 1:47
  • 1
    $\begingroup$ The volatility scales as the square root of time. So in one month, you would have $\sigma / \sqrt{12}$, not $\sigma / 12$. This has nothing to do with being log normal, though. $\endgroup$ – shabbychef Feb 1 '11 at 3:29
3
$\begingroup$

The distribution of the log of a stock price in n days is a normal distribution with mean of $\log(current_price)$ and standard deviation of $volatility*\sqrt(n/365.2425)$ if you're using calendar days, and assuming no dividends and 0% risk-free interest rate.

Note that the standard deviation is independent of the current_price: if $\log(current_price)$ increases by 0.3 (for example), the stock has increased by 35%, regardless of its current_price.

To include dividends and the risk-free interest rate, see:

http://en.wikipedia.org/wiki/Black-Scholes

which models future stock prices w/ an eye towards pricing options.

$\endgroup$
0
$\begingroup$

To create a lognormal distribution (that is, to generate values from it), you need to start with normally distributed numbers and then exponentiate them.

That is to say, take a sample $z$ from the standard normal distribution, and form the lognormally distributed underlying value

$$ U_T = U_0 \exp\left( (r-q-\sigma^2/2)T + \sigma \sqrt{T} z \right) $$

The probability density function of $U_T$ is formed from solving this for $z$ and then applying the normal PDF.

$\endgroup$
  • $\begingroup$ I get that $r-\sigma^2/2$ comes from ito calculus, but I don't get the extra variables $z$ and $q$, can you explain what that means or from where I can get that information? $\endgroup$ – silgon Jul 27 '17 at 13:13

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.