Hi,Can anyone tell me how to calibrate a SABR model when ITM vols and OTM vols are not quoted in the market?Thanks in advanceamkey

By definition of calibration if you dont have traded instruments you can't calibrate !That's true for any model.Then there are only two possibilities: 1-You have no market view2-You have a view If 1/==> then don't use a smile.if 2/ ==>Even if calibration of the smile is not possible and your view is that a flat smile is "unreasonable" then you can either use your own numbers or use another market that you think is close. But in any case you cannot call that calibration.

Thanks Clopinette for the answer. Is my first time calibrating market data to model. Guess I will use my own numbers for market without ITM and OTM quotes.

Hello,It is my firs time in a Wilmott´s forum. I am a very junior quant.Excuse my english but I am spanish and I do not speak english very well.I hope somebody here can help me.I am studing the SABR model, I am reading 'Managing Smile Risk' and my question is maybe quite simple for you:Can anyone of you explain me easily how can I find the parameters rho and volvol of the model??.Sorry if I am much too naive.Thank you in advance

Hi,Any links or book about SABR will be welcome !thousand thanks!

This paper here may have some useful answers: Calibrating SABR in Illiquid MarketsGood luck

There are three cases:a) you know what the market smile is, and then a weighted least squares (weighted because vols away from the ATM point are usually less well-known than near the money vols) sufficesb) other market participants know the smile, but you don't ... if you can find a friend to give you a snap shot of the smiles on a given day, you can fit the SABR parameters, and then hope that rho and volvol stay the same, even as the ATM vol changes; otherwise go to (c) belowc) no one knows the smile: in some currencies, the vol market is not developed enough for smiles to be agreed on. In this case there are too good guesses: i) using a CEV model with a guessed exponent (1/3?) ii) using a CEV model with different exponents for below the money strikes and above the money strikes. Popular is exponent 0 (normal model) for small strikes and exponent 1 (log normal model) for larger strikes1