Serving the Quantitative Finance Community

 
User avatar
figoliuxi
Topic Author
Posts: 11
Joined: August 23rd, 2011, 4:05 am

Question on Calculating KRD and durations: how to make sure curve consistency

March 21st, 2018, 8:44 pm

As we all know how to calculate KRD and effective duration: shock your curve (could be spot/forward/par) and get the price difference. 
However, there seems to be a problem with the method. If my curve was constructed from market instruments and now I would like to shock arbitrary month key point, I would basically need to add that point into my curve. However, after doing this, even my shock is 0, the new curve and old curve would not be exactly the same now. How can I claim that my KRD would be Pup - Pdn / P where P is the initial price?

More specifically, In the book by Andersen & Piterbarg, on page 253, they suggest the forward rate approach by solving dV/df, where f is forward curve. They suggest to have 3 months apart discretization grid. 

I think once you construct the discretization grid, your new forward curve would be different from your raw curve even though you keep interpolation the same since you are losing information.

Any thoughts on this?