Hi, I am wondering if I may pick your brains....Say we have a swap curve that comprises of deposits, FRAs/IRFs and swaps and we seek to build it the classical way (ie same discounting and projection curve). If I am to apply a shift of x bps, ie increment all quotes by x basis points and rebuild the curve, the curve would build for reasonable sizes of x. However, if I try to bump the curve by a lot, say 3000 bps or more, sometime the curve may not build. I am wondering if it is true that there is a limit on how much the curve can be bumped for any given curve? Or is it simply that the curve-building algorithm is not sufficiently robust?Thank you for your attention on this...

what is the Economic justification of Bumping a curve by 30% ?

Well....., the current context I have here is the calculation of Z-spread for bonds. If my understanding is correct, what we do is to solve for the x, such that, if we shift the swap curve by x basis points, rebuild it, and then PV the bond off the curve, it will bring it to the current dirty price. If we have a high yield bond with low value, it is entirely possible that we may need to bump a curve by a large amount in calculating the z-spread. Evidently we can get round this by keeping the original swap curve intact, but supplement the discount factors with a spread, which would give us similiar results...But having said this, it is interesting that, for example, the USD swap curve has a bump limit of around 1500bps (give or take a few hundreds) beyond which, the bumped curve will not build. What appears to have happened is that the fixed leg coupon payments gets too large that we need negative discount factors at the later timepoints to bring the fixed leg back to par.... Has anyone similiar experience on this?

QuoteOriginally posted by: Samsaveelwhat is the Economic justification of Bumping a curve by 30% ?Apocalyptic CCAR?

Don't shift inputs to the curve generator, shift the curve itself. See http://www.wilmott.com/messageview.cfm? ... SGDBTABLE=

coming back to the original question, I think there is no limit in absolute terms (a flat yield term structure with a continously compounded zeroy yield of y is perfectly valid and arbitrage free for each y > 0). However there are restrictions on the shape of the curve. Say you have already stripped your curve until some t_{n-1} and you wish to add t_n by considering a swap maturitng at t_n (one curve setup, very simple): The condition is (for t_s = settlement, s = swap rate)[$]1.0 - s \sum_{i=1}^n \tau_i P(t_s,t_i) = 0[$]which means in terms of zero yields[$]e^{-r_n t_n} = (1.0 - s \sum_{i=1}^{n-1} e^{ -r_i t_i }) / \tau_n[$]therefore[$]r_n = - \ln ( (1.0 - s A)/\tau_n ) / t_n[$]which gives the upper bound for the swap rate s, dependent on the already stripped points[$] s A < 1.0 [$]with [$] A = \sum_{i=1}^{n-1} e^{ -r_i t_i } \tau_i [$]where A of course can get arbitrary small.