[Quantlib-users] Amortizing swaptions

[Aleksis A. R. <ale...@go...>]

Hi. When playing around with quantlib, it seems using the 1-factor GSR model allows computation of swaptions on amortizing swaps by simply supplying a standard swaption vol cube. Just wanted to get some feedback on the veracity of this approach because these swaptions rely on correlations between fwds in the amortization profile (not supplied in this model, so really they require a 2 factor model to price). In the example I compared a 1y10y vanilla swaption to a 1y10y swaption with constant (linear) amortization profile, both with their respective atmf. The numbers do make sense (vega, delta and premium is around half of the vanilla), however the vega risk profile of the latter seems to bucket the risks to the 1y6 and 1y7y buckets (ignoring 1y2y,1y3y,1y4y…etc which you would expect to see):

 
1Y	6Y	7Y	8Y	9Y	10Y	Total
1M	7	0	0	0	0	0	7
1Y	0	7,617	10,665	(5)	(7)	(4)	18,267
1Y6M	0	90	121	(3)	0	0	208
Total	7	7,707	10,786	(8)	(7)	(4)	18,482

Thanks, Aleksis