Key rate durations

[rhoucag]

I've been doing key rate duration analysis on our company's assets lately. It has occured to me that giving the duration for a specific term is useless unless you know the other terms choosen as well.

For example, suppose you set the key rate terms as .25, .5, and 10 years. Suppose all durations happen to equal 1.0. Wouldn't the 1.0 mean more for the .25 term since the difference in adjacent terms is only .25 years ? The 10 year value of 1.0 would indicate very little sensitivity to the long term rates since the 10 year key rate is showing the affect of all rates above and a portion of all rates between .5 and 10.

Am I correct in my gut feelings ?

[An Affair to Remember]

k r d is a partial derivative at a given point.

Mathematically, sum of k r d's = duration.

[NoName]

1. Let's say that you decided to do key rate durations for terms of 0.25, 0.5, 1, 1.99999, 2, 2.00001, 5, and 10 years. If what you're saying is that it would then be very difficult to get a high KRD for the 2, I think that's correct.
2. I don't think a single number ever means anything in looking at KRDs. You have to look at the whole set. What purpose would you be using KRDs for that would provide a context where a single KRD number is meaningful?

[rhoucag]

Thanks for the responses.

I don't see a use for looking at one KRD alone. I just wanted to see if my feeling was correct.

I also feel that you can only compare different KRDs equally only if they are equally spaced apart, i.e. 2, 4, 6, 8, ...