Implied Volatility vs. Historical Volatility

[JPslx]

Hope I posted this under the right section!

I am a new participant to Bionic Turtle, and I have to say it's been quite educational.
At the moment I am developing a trading system on Foreign Exchange, and I am a little puzzled on how to build my risk management device. When investing spot, the maximum loss can be decided by placing a stop loss order at the appropriate level.

The challenge is finding the "appropriate" level: market wisdom would have you pick a technical level that the market "should not" touch unless the current trend fails; instead, risk management would have you calculate a VaR based on the volatility of the currency cross analyzed.

My question is regarding the appropriate measure of volatility to use. I am inclined to use a Garch(1,1) as it is renown to capture volatility smiles most adequately. Do you have any other suggestions? Also, for retreiving FX Option quotes is there a good free source of information? I would love to analyze the differences in projections using Implied Vol vs. Historical Vol...but I haven't been able to retreive useful option data...

Thank you for any answer.

 

[David Harper CFA FRM]

JPSlx,

Thanks for visiting. This is a specialized area, where I do not specialize, so apologies I can't really help.

But it's funny you asked because one of my favorite blogs published this earlier in the week:
http://www.voxeu.org/index.php?q=node/876
(I am still reading the underlying paper to learn more myself, it is a tough read). You might find this interesting because in addition to GARCH(1,1) they consider two other models. I don't quite understand the headline yet; i.e., why *exactly* they claim predictive ability...I hope that helps...I'll blog on this soon but i'll need to understand it first

Correct me if i am wrong (I ask b/c I am uncertain), GARCH(1,1) is better, as you say, at getting the smile merely (and only) because it is conditional (i.e., time varies). The FX is known, like many asset returns to be fat-tailed (leptokurtotic) and GARCH(1,1) can get at that, but really only because it conditions on the latest information? In other words, it is a time-varying mean/variance that produces a fat tail effect and I'd assume (perhaps) with the smile. I am just curious b/c I assume many of these GARCH "extensions" for FX do something more direct? Thanks, David

 

[JPslx]

Thank you for the timely respone...and forgive me for the absence (been away from the computer preparing for an exam).

About the "why garch(1,1) works"...it may be the italian blood in me that makes me affine to wacky ideas, but I believe that:
- the accademic work done to analyze arch effects is fantastic...but we may have gone beyond what's actually usefulness. I believe (from a very humbe point, since I'm not even a Masters in Finance yet) that accademic literature is nowadays overoptimized. I believe that the basic mecchanisms that describe time series and volatilty in particular have been "uncovered", and that general models like the Garch(1,1) are able to capture a wide number of inputs..it's just like a Trading Algorithm: the more "barriers" you give it before a signal can be generated, the more you limit it's degrees of freedom. Since the Market is an irrationally rational place, the best instruments tend to be (especially in FX) the generalized ones. That's why I believe Garch(1,1) works: in it's limited specificity, it's general enough to capture visible and non visible relations.

- the market structure is long from being efficient. The EMH (and this is comming from a CFA candidate) is appealing in theory because of it's quantitative nature, but absolutely irrealistic (see Stiglitz, 1980 for a sexy explanation). My approach to trying to understand market dynamics is closer to the Adaptive Maket Approach (see Andrew Lo, 2004). I believe that we all appeal to quantitative work (and go to extremes) because of the detail that gives us a false sense of security about what's going on. But it happens that, just like in a chess game, the more one complicates, the tougher it is to draw back to the big picture (i.e. putting it all together to come out with 1 decision). Therefore, my approach follows the Adaptive KISS rule: keep it as simple as the evolution of market dynamics (and knowledge) allow.

I will now proceed to read the article in the link you posted, for which I am grateful.

I hope my answer isn't too qualitative in nature, it's just that I've gotten nausea from frequenting the Willmott forum :-P

Cheers,
JPSlx