That article treats a single, pretty extreme, equity example. In any event, if I worked in interest rate space, I wouldcertainly try it out, as it is 'model-free'. (i.e., stochastic process free) (Still hoping for the little tutorial I mentioned in my Tue Jul 03, 12 04:55 PM post, from anybody)

I see this model more as an interpolator and smoothing method for illiquid markets. I used it a while ago for energy options. Starting from prices, potentially with arbitrages, Fengler's method gives you a smooth implied vol, arbitrage free. Then you can plug it into your favorire dynamics model: which gives you the risk parameters.

it is definitely used for derivatives pricing. what are you talking about? what about single observation multiple-index deerivatives? also people have tried and are trying hard using it for hedging purposes.other than that you are not really saying anything more than what I have said... QuoteOriginally posted by: miquantVery interesting conversation, even if it is obvious that nothing serious has been proposed so far to change the way volatility smiles for swaptions and cap/floors market can be reproduced.It is very important to notice (and understand) that SABR model is only used for smile calibration purpose, I know nobody using it for pricing or hedging any derivatives in IR. But it is of crucial importance since the volatility smile is what is used to get price of calibration products, being swaptions or cap floors for a BGM/HJM or so, being digital caps for Markov Functional model, but also to feed integrals in the CMS derivative pricing.In my opinion, that is why it is very important to get a smile with no arbitrage...I have been myself thinking for a long time that SABR is not really a model in the practice but just an super-interpolator, as such as any calibrated model used to price derivatives. But lately I have a very different feeling and understanding.I think it is very difficult to trace the limit between model/interpolator in SABR case. One one hand, we want it to pass through a number of points (those are market vols), and then we need to be able to use it for any other point not in the market. Until recently it was all fine thinking of SABR as an interpolator because it was acting just fine around the ATM and it was all we needed. The problem came with lower forwards or other related problems that brought the need to get accurate prices on the left wing. Problem is that SABR doesn't behave good on that part of the smile and now it is revelant to think that the problem comes from the model limitation. So yes SABR is more than an interpolator now, and yes it needs alternatives to take care of what happens in the market.I am not giving any solutions right now, but my point of view for the one who are not that familiar with the use of SABR for practicians.

The SABR model gives you all this and if you modify a bit it will fulfill all points you list.QuoteOriginally posted by: ZhuLiAnYes you extrapolate for high/low strikes and this introduce a model risk for low strikes swaptions, CMS and CMS spreads. A first step toward a SABR alternative would be to list what are the exact requirement of such a model: - The model is arbitrage-free - The model calibrates the swaptions market. This means that it should produce a SABR Hagan like skew/smile near the ATM region. - The model is parsimonious and the parameters are intuitive. Ideally the parameters should have orthogonal effects on the implied volatility surface. - The extrapolated volatility should imply realistic left and right wings since this give prices for low strike options and CMS products. - The model should allow fast pricing because of its used in the calibration of the swaptions volatility cube. - The model dynamics should produce robust and realistic hedges. Anything missing?

I agree and I don't know. I think that this is were very practical aspects cross fundamental aspects. Remember that Hagan et al had the experience to actually write the SABR expansion as two impled vol expansions, one normal and one lognormal. They could have stopped with the pricing formula.I guess that the reason for why people prefer a parametric smile model over a smile curve is a bit the same. They kind of know that all these asymptoticexpansions are quite limited, yet having them in a system giving you a fully fledged set of greeks is convenient. and there is a territory (i.e. gamma) where the greeks are actually potentially meaningful. Then there is the fact that you don't really want to stray away too far from what other people are using. I mean vanillas are bid-offered outside of any model. Imagine you use a real fat-tail model to manage your shop, soon you will be no more.I wold turn the question around and wonder why equity people are happy with some of the models they use. That being said, I know that some are not and finding a decent backbone model in equity land is a holy grail too.QuoteOriginally posted by: AlanWell, in equities, there is a popular model-free arb-free smile fit due to Fengler. Why not adopt that with some arb-free extrapolation scheme?

I can't give you a tutorial, I can write down some points. I think that the fixed-income sell-side is at sea completely for far OTM strikes. I doubt very much that other asset-classes are dong better. The problem is that the option skew reflects the option regime of the future (not to speak of the unerlying regime of the future) and nobody knows what that is. In fact I don't think that market makers even spend much time thinking about this. So anything that is a few % away from the ATM, who knows. On the buy side, the perspective is a bit different. You cannot see the flow, but you can poll dealers so you can piece together a picture from dealer quotes, quotes of CMS products (which contain some information about the smile) and then macroeconomic fundamentals (very important). I raised this point before in other threads.Here another little rant. I personally think that finance is very problematic, because as opposed to the natural sciences where you actually get a chance, if you want to, to do empirical work, it is called experiment there, in finance it is really quite complicated to get that access. In finance experimentalists are called traders (and you have to quite carefully distinguish sell side traders (market makers) from buy side traders (arbitrageurs, speculators, punters, etc...). Theoreticians are called quants. Most theoreticians in finance (me first of all) don't really understand the experimental side of the field. A lot of this has to do with the fact that financial data is hard to observe. It becomes a full-time job. You need to really observe the market dyanmically, i.e. intra-day.If you look at it close to close, there is only so much you can learn. Further, it is tricky to conduct an isolated idealized table-top experiment like you can in physics. Maybe it resembles more biology in this sense. Moreover experimentalists (traders) keep an impossible barrier between them and the rest. It is all about edge, competitive advantage, prorpietary information, etc... The upshot is that most quants, I really mean most of them, even the very good ones who write the fat volumes, view quantitative finance as a collection of static facts and methods, like day count conventions, model families, etc... All in all one of the reasons why progress is so slow. It is just not in people with power interests.QuoteOriginally posted by: AlanThat article treats a single, pretty extreme, equity example. In any event, if I worked in interest rate space, I wouldcertainly try it out, as it is 'model-free'. (i.e., stochastic process free) (Still hoping for the little tutorial I mentioned in my Tue Jul 03, 12 04:55 PM post, from anybody)

QuoteOriginally posted by: mtsmI can't give you a tutorial, I can write down some points. I think that the fixed-income sell-side is at sea completely for far OTM strikes. I doubt very much that other asset-classes are dong better. The problem is that the option skew reflects the option regime of the future (not to speak of the unerlying regime of the future) and nobody knows what that is. In fact I don't think that market makers even spend much time thinking about this. So anything that is a few % away from the ATM, who knows. On the buy side, the perspective is a bit different. You cannot see the flow, but you can poll dealers so you can piece together a picture from dealer quotes, quotes of CMS products (which contain some information about the smile) and then macroeconomic fundamentals (very important). I raised this point before in other threads.Here another little rant. I personally think that finance is very problematic, because as opposed to the natural sciences where you actually get a chance, if you want to, to do empirical work, it is called experiment there, in finance it is really quite complicated to get that access. In finance experimentalists are called traders (and you have to quite carefully distinguish sell side traders (market makers) from buy side traders (arbitrageurs, speculators, punters, etc...). Theoreticians are called quants. Most theoreticians in finance (me first of all) don't really understand the experimental side of the field. A lot of this has to do with the fact that financial data is hard to observe. It becomes a full-time job. You need to really observe the market dyanmically, i.e. intra-day.If you look at it close to close, there is only so much you can learn. Further, it is tricky to conduct an isolated idealized table-top experiment like you can in physics. Maybe it resembles more biology in this sense. Moreover experimentalists (traders) keep an impossible barrier between them and the rest. It is all about edge, competitive advantage, prorpietary information, etc... The upshot is that most quants, I really mean most of them, even the very good ones who write the fat volumes, view quantitative finance as a collection of static facts and methods, like day count conventions, model families, etc... All in all one of the reasons why progress is so slow. It is just not in people with power interests.QuoteOriginally posted by: AlanThat article treats a single, pretty extreme, equity example. In any event, if I worked in interest rate space, I wouldcertainly try it out, as it is 'model-free'. (i.e., stochastic process free) (Still hoping for the little tutorial I mentioned in my Tue Jul 03, 12 04:55 PM post, from anybody)Thanks for the response. I agree completely re intra-day observation. In single name equity, half the game isthe after-hours action on earnings release day. Yet the academic literature almost (not quite) ignores most of this.

QuoteOriginally posted by: mtsmThe SABR model gives you all this and if you modify a bit it will fulfill all points you list.QuoteOriginally posted by: ZhuLiAnYes you extrapolate for high/low strikes and this introduce a model risk for low strikes swaptions, CMS and CMS spreads. A first step toward a SABR alternative would be to list what are the exact requirement of such a model: - The model is arbitrage-free - The model calibrates the swaptions market. This means that it should produce a SABR Hagan like skew/smile near the ATM region. - The model is parsimonious and the parameters are intuitive. Ideally the parameters should have orthogonal effects on the implied volatility surface. - The extrapolated volatility should imply realistic left and right wings since this give prices for low strike options and CMS products. - The model should allow fast pricing because of its used in the calibration of the swaptions volatility cube. - The model dynamics should produce robust and realistic hedges. Anything missing?The initial question is about this modification. Adding a shift is not enough because the Hagan normal vol expansion is very extreme for high strikes.

Could you please post a smile chart that illustrates your issue? Thanks.

QuoteOriginally posted by: mtsmit is definitely used for derivatives pricing. what are you talking about? what about single observation multiple-index deerivatives? also people have tried and are trying hard using it for hedging purposes.other than that you are not really saying anything more than what I have said... Single observation Multiple Index derivatives, what the hell is that, example please?? Is is an IR product??Instead of being a smartass, you should read my post again mstm: I said that I know nobody using SABR for anything else than mkt volatility fitting and that's the case ... you say "people have tried and are trying..." so pragmaticaly speaking, SABR is not used currently in Front Office for other purposes.Can you explain how people wants to use it for hedging any exotic product, I am not sure I understood that point??

miquant, you are being hereby ignored as you clearly do not know what you are talking about.zhulian, I am very interested in the negative rates extension, it keeps coming up again and again.I have a comment to make about your comment on the normal expansion. can you pm me please?QuoteOriginally posted by: ZhuLiAnQuoteOriginally posted by: mtsmThe SABR model gives you all this and if you modify a bit it will fulfill all points you list.QuoteOriginally posted by: ZhuLiAnYes you extrapolate for high/low strikes and this introduce a model risk for low strikes swaptions, CMS and CMS spreads. A first step toward a SABR alternative would be to list what are the exact requirement of such a model: - The model is arbitrage-free - The model calibrates the swaptions market. This means that it should produce a SABR Hagan like skew/smile near the ATM region. - The model is parsimonious and the parameters are intuitive. Ideally the parameters should have orthogonal effects on the implied volatility surface. - The extrapolated volatility should imply realistic left and right wings since this give prices for low strike options and CMS products. - The model should allow fast pricing because of its used in the calibration of the swaptions volatility cube. - The model dynamics should produce robust and realistic hedges. Anything missing?The initial question is about this modification. Adding a shift is not enough because the Hagan normal vol expansion is very extreme for high strikes.

QuoteOriginally posted by: mtsmmiquant, you are being hereby ignored as you clearly do not know what you are talking about.And you being a smartass again!!! Are you a trader?