I have implemented and calibrate the SABR model. It works well. The problem is that the SABR model only works with short dated options, <4Y. However, I would like to get volatilities for long dated options. >20YWith the Sabr model, for long dated option, I find negative volatilities!! Bloomberg displays the volatilities for swaptions at any matutities. I don't know how they get those volatilities. What do you think of the Heston's model;

if you don't match the vols on long dated option vols from bloomberg, do you not think that it's possible you haven't calibrated your model properly?

I totally agree with you. However, theoritically, the SABR model is not efficient for long dated options, what ever the calibration might be.

> However, theoritically, the SABR model is not efficient for long dated options, what ever the calibration might be.How did you arrive at this conclusion?In my experience, I didn't have troubles with long maturities.

Gatheral does state that it won't work for long maturities because volatility isn't mean reverting, but i agree, seems to be simple to calibrate for long maturities

QuoteOriginally posted by: WibbleGatheral does state that it won't work for long maturities because volatility isn't mean reverting, but i agree, seems to be simple to calibrate for long maturitiesI don't think it is a good idea to use SABR as a global (dynamical) model, that is to calibrate it to a few maturities simultaneously. The SABR model is typically used to parametrisize implied vols at a given maturity time. Once you have decided to use the SABR to explain the skew/smile at a single point in time (assuming it is a static model), you don't have to worry whether vol is mean-reverting or not - you are just dealing with its terminal distribution at the given maturity time.

How did you arrive at this conclusion?>>I didnt: But some specialists did. For some theoritical reasons: The sabr model is derived from small perturbations. So it only works for shot maturities, with strike close from the forward.