Does Nelson-Siegel require adjustments to yield curve input data?

Question

I am attempting to gain a better understanding of the limitations of the Nelson-Siegel model as described in Estimating the Yield Curve Using the Nelson-Siegel Model.

As I have been playing around with the data I started to wonder whether the inputs to the Nelson-Siegel model are correct. I am using Daily Treasury YieldCurve Rates and estimating the model through the R YieldCurve package. It has been my understanding that spot rates need to be derived from observable par yields before applying any modelling. This understanding, I believe, has been confirmed at a separate discussion. But documentation of the relevant R packages fails to mention which rates should be supplied.

Should the input to the Nelson-Siegel model, in general and with respect to the R package, be the Daily Treasury YieldCurve Rates or should one bootstrap the spot rates before applying the model?

Answer 1

The NS model should be fit directly to bond prices. If you have the prices of all the Treasuries, you should use those directly. See this paper for how the Fed does it http://www.federalreserve.gov/pubs/feds/2006/200628/200628pap.pdf

The "Daily Treasury Yield Curve Rates" are already fitted par yields (they're fitted using a cubic spline model to on-the-run and select off-the-run Treasuries). Since these are par yields, you can assume that they represent the coupon rates of Treasuries with prices of 100, and feed them into the NS model.