August 14th, 2005, 11:16 am
Let's assume it's a forward starting bond. In that case, when you try to compute the duration numerically by bumping the curve by 1bp in parallel, every projected cashflow will be increased by 1bp. However, you will discount these higher cashflows at a higher rate and it turns out that the net effect on your bond price is zero. However, once the first cash flow has fixed, the above doesn't hold anymore. The first cash flow doesn't change anymore, so when you bump the curve by 1bp the discount factors will change, so the net effect will be zero for the cashflows which are projected but haven't fixed but non-zero for that first cash flow. The difference in PV will be roughly equal to the daycount fraction to the first cash flow which will be equal or less than the reset frequency.
Last edited by
Geist on August 13th, 2005, 10:00 pm, edited 1 time in total.