Historical simulation, value at risk (VaR)

This is an illustration of historical simulation using a single-asset (versus a portfolio). The asset is Google's stock; I pulled daily (periodic) returns for the last 100 days. Note the histogram exhibits fat or heavy-tails with the two outliers including Friday's 10% drop. This example shows some of the pros and and cons of historical simulation: Easy to use (pro): it is just a percentile or quantile function; Unlike parametric or Monte Carlo, no model risk (pro); Treats all historical observations equally (con); May not be enough data in the tails (con, a classic challenge)

Category:

Education

Tags:

• Finance
• risk
• value
• at
• statistical
• distribution

[ruede80]

I don't get the formula to calculate the percentage change from 17th to 18th. Why is he using the ln?? LN(B8/B9). Is this a regular formula??

If day one close was 10 $ and day two close was 12 $ I would calculate (12-10) / 10 that would be 2/10 = 0,2 = 20%.

After the formula using the LN it would be LN(12/10) = 0,1823 = 18,23 %

Can anyone explain me that???

[bionicturtledotcom]

@ruede80 your way is not wrong, it is the simple return: 20% = (20 -10)/10. I have shown the continuously compounded return (aka., the log return), such that 10*exp(18.23%) = 12. Continuous compounding is what you get if you keep increasing the frequency, from annual to daily to intra-daily to continuously. It has some nice properties, mainly that you can add continuous returns over time easily.

[JJFZ3000]

@khanpreston1

That's just convention.

[khanpreston1]

y confidence level is always 95% or 99%?how is it determined?

[nellykrac]

what is the formula to sort. from the 2:01 to 2:03

[dayflynn]

thanks a lot

[jmatheus]

nice...keep on rockin!