Motivated by a discussion on the Webring
I've always been amused (well, annoyed, actually) when somebuddy uses standard deviation to measure "risk".
>Are we doing that again!
Yes! It bugs me!
Consider two assets, A and B where:
(the returns of asset B) = (the returns of asset A) + 10%
>You just add 10% to A's returns?
Yes, so which asset is "riskier"?
>I have no idea, but I suspect it's asset A.
Actually, if you use Risk = Standard Deviation, they're equally risky.
>Do you have a picture?
Patience. Anyway, what I'm saying is that adding a constant to the returns doesn't change the standard deviation and, in my opinion, that makes standard deviation a lousy measure of "risk".
>Well ... maybe "risk" measures uncertainty, eh?
Okay, consider assets A and C, where:
The returns for asset A oscillate between +5% and 5%.
The returns for asset C are fixed at 5%,
except that, every third year (as a Christmas Bonus )
you get extra shares so your return for that third year is 12%.
>So you get 5% then 5% then 12% and that repeats, right?
Right, so which asset is "riskier"?
>Asset C is some stock!
Actually, it's a mutual fund called the Mickey Mouse Fund. It gives a 7.3% average return over three years.
Anyway, which asset is "riskier", A or C?
>I have no idea, but I suspect it's asset A.
Actually, if you use Risk = Standard Deviation, asset C may be riskier.
>Do you have a picture?
Yes, here's an example. Now tell me whether asset C is more "uncertain" than assets A or B:
>Is that it?
Yes, that's it. I just wish they'd call it standard deviation or maybe volatility ... but not risk
Standard deviation is just a convenient measure of how far returns deviate from their average.
Big standard deviations need NOT mean a big probability of a loss ... or even big uncertainty.
>But risk, defined as standard deviation, is financial technobabble. Why should it be Webster's "risk"?
Okay. Good point. So if the gurus kept it to themselves, I'd be happy.
>You needed to get that off your chest, eh?
Yes, I feel much better already
A poll conducted on the Webring Forum gave this result
>Over a thousand responses to the poll question, right? That'd make it statistically significant, right?


Well, close to a thousand ...
>The percentages don't add to 100%, so I guess they just take the integer part of the percentage?
Maybe.
>So the number of responses could be 200, 1100 and 400.
Maybe.
>Or they could be 2, 11 and 4.
Maybe.
