Implied Volatility Surface Dynamics and rules of thumb - does it matter? Please help!

[rriboldi]

Hi to everybody,

I have a question for those of you who worked in banks or do know what market practices are with a certain degree of confidence.

I read Derman's 1999 paper on regimes of volatility, followed by Carol Alezander's one on PCA and implied volatility smile and skews. Given the fact that we know that there are these rule of thumbs and that we can test them, conditionally on knowing in which market regime we sit in (and also here, how does one decide?), does this really matter?
I mean: I had an idea of trying to replicate Derman's work by extending the application not only to 3-months options but also for other maturities and trying to have a look at how the surface moves not daily, but also with different frequencys for a university research. The point however is: in the markets, and specifically from the point of view of the trader or the risk manager how could such type of information be useful?? Do the usage of more complex models such as stochastich volatility, jump diffusion (or a combination of both) actually remove the interest in knowing according to which "rule" the volatility surface moves, since they already "per se" imply an evolution for it?
What I mean basically is: once a desk has decided which of the many models currently out there to use, do all the reasonings which could be derived by a Derman/alexander type of analysis matter?
If so, could you please explain, practically, what the use (if any) of such findings would be for the risk manager/trader? (this is to understand whether it makes sense or not to do it)
Otherwise, what other type of analysis could be done on the typical evolution of the volatility surface without having to "choose" one of the models, while keeping the results applicable and of interest for option desks?
In all this it isobvious that one has also to clarify which type of desks could benefit from such results, since obviously exotic desks differ from plain vanilla desks in many ways, and the same can be said for differences in trades trading different types of maturities or OTC vs traded options. It would be nice to understand what type of information is suited to what type of desk, and I would be very grateful to any experienced member who would like to share his opinion.

Would otherwise a comparison between different models (such as Ledoit-SantaClara vs Schoenbucher vs Heston etc) be more interesting? In that case however one would need to understand which one is the most used on the market.


Thanks in advance

 

[Hoggums]

Yikes. My head hurts after reading that.

you'd be better off asking some quants here-> Wilmott | Serving The Quantitative Finance Community | Forums

 

[rriboldi]

Yikes. My head hurts after reading that.

you'd be better off asking some quants here-> Wilmott | Serving The Quantitative Finance Community | Forums

I know that forum, however what I would like is TRADER'S opinions about this matter, since I wanted to deliver a study which could be of use on volatility desks.

I'm sure that there are lots of option traders who have a view on this topic.

 

[MrMiyagi]

I know that forum, however what I would like is TRADER'S opinions about this matter, since I wanted to deliver a study which could be of use on volatility desks.

I'm sure that there are lots of option traders who have a view on this topic.

try the guys over at fattail.org there's a hardcore of successful option trader/quants over there that could help you out.

You should still post on wilmott, there's quite a few big vol/stat arb traderquants who post over there... realistically its traders on bank desks and sophisticated hedge funds that you need... (they'll ignore your post though if the answer to your question can easily be sought and found in their archives) oh and Nuclear Phynance as well

 

[slik]

There is no obvious answer to what you ask- the vol surface is a pretty arbitrary thing that is pushed up and down by supply and demand- if a big trade goes down in a market, it can flip call skew to put skew without any underlying mathematical reason other than the fact someone likes calls/puts - which incidently doesn't always suggest a liking for the direction of a trade (e.g. a bond holder may sell calls on a rally not because they are bearish but they are hedging a long bond position).

To use stoichastic analysis in vol prediction is pretty much the same as using such analysis in prediction in futures/stock movement- how much faith you put in it depends on 1) The amount of capital you have, the market can remain insane longer than you solvent and 2)your assessment of market conditions and how much a black swan can happen..... to give you an idea in early 2007, 30 day Shatz (2 year bond yield on German government bonds) option vol traded 0.7%, sold down from 1.7% in the last 12 months, so stoichastics would say trend downwards - sell the living snot out of it, then China equities wobbled in Easter 2007, vol spikes to 2.2% (triples!), trades between 1.2- 2% until 2008, then during the last year and events spikes the vol to 4% during the Lehmans debacle.

My point is this thing is as difficult as predicting the path of the underlier and a true market observation such as a view a market maker would have is fit the bugger to market, there is no amount of stats or analysis to truely, and accurately model a random event, let alone one with big ass rip your face off fat tails.

 

[slik]

To give you a rock solid example of skew (calls/puts trading over each other in terms of vol), implied volatility in soverign Bonds for the last 4 months have had a call skew- reason, flight to quality, as equities ticks off or dumps, panic arises and people buy bonds (or calls as a long delta buy proxy) so you can understand vol or options being bid on a rally with bonds (fine as it happens as models can be adapted and you can increase vol automatically on a rally - a parameter called VCR or volatility change rate), however in the last month or so, puts are getting better bid leading the vol across a month to resemble a true smile - why is this?

There is nothing mathematical about it, it is because the economics of a loose fiscal policy ala Bernanke/King pumping money into the system has to lead to in the end to inflation.... maybe not now but in 6/12/24 months, at some point inflation will rear its head and bond prices will make the equity sell off look like a walk in the park.

The only heuristic that I can give through trading options for the last 5 years is that when there is a panic in the market, the skew swings in ways that is impossible to model and that the term structure inverts- in calm periods such as 2005/6, back month vol always traded over front month vol. In volatile markets, front month vol usually trades over back month vol- this makes common sense in that the panic is usually a reaction to a recent event/news that will move the spot violently in the near term and therefore gamma is bid. Where is gamma the most in bang for your buck options? front month/short dated options.

 

[rriboldi]

Thanks a lot for the answers.

Now,

could someone point out where one can find historical data on SPX or FTSE/DAX options (at least close and volume) which is less costly than the MarketDataExpress (got here through the CBOE website) or the LIFFE website? I could also get the hands on a Bloomberg terminal, but I don't know whether those quotes are reliable and how much historical data I can get from there.

 

[slik]

Option data is not easy to get. You can get some from bloomberg, but implied vols are not stored for very long on bloomberg. It is not easy to analyse this kind of data either; you need to compare 30 day options with 30 day options etc

I remember my firm buying the data directly from the exchanges. I wouldn't bother buying all 3 of the equity indices, just buy one, the others will look pretty much the same in these interconnected markets.

 

[egro1egro]

Paul Wilmott is very pessimistic about implied vol models and said that even inventors of this model are not very keen on it. Paul says that the moment the publication went out - is one of the "lowest" moments in quant finance. Paul's argument is that the model is bad because reasults are unstable. Today you see implied vol for a certain option high, tomorrow for the same option - low. Implied vol doesn't predict any volatility at any price level/maturity, doesn't give the right option prices (particularly for exotics) and doesn't reduce risk.
Heston model is difficult to calibrate (with its 999 parameters-)) and calibration doesn't mean the right price either.
But both approaches are very popular and widely used in banks regardless.