November 4th, 2012, 8:47 am
QuoteOriginally posted by: prodiptag If we have an ATMF midcurve swaption straddle position, one way to replicate this is via a long and short straddle, the strikes being the respective ATMF strikes. What if the original midcurve is quite OTM, any idea how to split the strikes of the long/short straddles combo - preferably in as much model independent way as possible?What do you mean by "midcurve swaption"? Is this by extension of the STIR future jargon an option on a swap that starts 1 (or 2) years after its expiry?If this is the case, it can not be treated as long (the full period) and short (the period up to start). That is correct at the swap level, not at the swaption level. The basket of option (long/short options) is not equal to the option on a basket (midcurve swaption). The long/short options give you an upper bound on the price of your midcurve, not an unbiased estimate of the price itself.The only way I know off to price this type of deal is to use a multi-factor model and to calibrate it to the two option you mentioned. It is important to use a multi-factor model, as described in [Hunt and Kennedy. Financial Derivatives in Theory and Practice. Section 15.4], as a single factor model will always be an upper bound for the price. Often those options are treated as special cases of non-constant notional swaption. Calibrating to the two options you mentioned gives you hedge ratios with them, but the ratios are dynamic, not static.This does not really answer your question on how to split the strikes. The answer to this question is also required as an input to the calibration basket.You can also use a multi-factor with smile model (like LMM-SABR) and you don't have the question of the calibration, as the model "calibrate the full market", but you still have the question of the hedging instruments.I hope this help.