Nelson–Siegel Curves for Key Rate Sensitivity Analysis

Summary

The document asks what interpolation method is suitable when perturbing a key interest rate to estimate the sensitivity of a portfolio’s value. The question contrasts a book’s use of linear interpolation with the concern that a linear curve may be unrealistic, especially for long-duration portfolios or short-term perturbations. The answer proposes the Nelson–Siegel model as a possible way to represent the yield curve.

The reply is only a brief suggestion and provides no calibration procedure, equations, comparisons, or evidence that Nelson–Siegel is the standard choice for key rate perturbations. It also does not explain how to construct local curve shocks or preserve consistency across maturities. The material serves as a pointer toward a parametric yield-curve model, not as a complete recommendation; practitioners would need to assess the model’s fit and the sensitivity of their results to the chosen curve representation.

Key ideas

• Key rate perturbations require a choice of how rates between maturities are interpolated.
• The question raises concern that linear interpolation may not represent curve behavior realistically.
• The response suggests the Nelson–Siegel model as a possible yield-curve representation.
• The excerpt provides no implementation details or evidence that this method is standard for sensitivity analysis.

Tags

# What type of interpolation should be used in key rate perturbation models?


# What type of interpolation should be used in key rate perturbation models?












When perturbing a key rate in order to assess sensitivity of portfolio value, what sort of interpolation is standard? A book I am looking at says linear, but this seems pretty unrealistic to me--and of real significance for longer durated portfolios or shorter term perturbations. Anyone familiar with literature on the matter? Thanks.

## Answer by Lior Yochpaz (score -1)

https://quant.stackexchange.com/a/27927

If I have understand right what you are looking for, I think that Nelson–Siegel Model can halp you, google it.

Hear some readng: Nelson-Siegel model

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.