Choosing Yield Curve Levels or Daily Changes for PCA

Summary

The document compares principal component analysis applied to yield curve levels with PCA applied to daily rate changes. The researcher reports that both input choices produce component shapes resembling a parallel level shift, a slope change, and a butterfly movement. For daily changes, the researcher leaves the data uncentered and unstandardized, reasoning that changes should already be centered near zero, while asking whether levels should be centered or standardized because they may contain drift.

The only response says the choice depends on the intended simulation: rate levels may suit simulations of rate increases, while changes in slope may suit a slope-focused objective. The exchange offers no comparative tests, preprocessing guidance, or evidence that either input is generally preferable. It frames PCA choice as dependent on the modeling goal and leaves questions about centering, scaling, and level-versus-change inputs unresolved.

Key ideas

• PCA on yield curve levels and daily changes can both produce level, slope, and butterfly component shapes.
• The choice of input depends on the intended use of the principal components.
• The discussion raises centering and standardization questions but does not resolve them.
• No comparative statistical tests or evidence establish a generally preferred input.

Tags

# Yield curve PCA: levels or daily moves?


# Yield curve PCA: levels or daily moves?












I have tried using both yield curve levels as well as daily moves (absolute change) while doing PCA. Using both types of input/dataset gives me roughly the same shape in terms of principal components matching the empirical observation: parallel shift, slope, butterfly.

For daily moves, I am not doing any centering nor standardization since it should already be centered at 0. For spot rates, given that there's a drift component, are there any preferences into doing either centering or standardization?

I am curious if there are any preferences between using spot levels vs. daily moves as well or are both acceptable. Since I've seen most people doing it on the spot levels only.

Thanks!

## Answer by Kevin (score 0)

https://quant.stackexchange.com/a/70456

I think it depends on whether you're trying to simulate an increase in rates (level) or increase in slope.

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.