We use a proprietary dataset of daily quotations of individual bonds from the period 2005–2018 to assess the appropriateness of three common yield curve intrapolation methods: Nelson-Siegel (NS), Diebold-Li (DL) and cubic splines. Spot (zero-coupon) yield curves derived from the methods are used to price all securities in the sample in both in-sample and out-of-sample settings. The results highlight trade-offs inherent in any empirical strategy: between flexibility and in-sample fit, between overall fit and fit in particular segments of the curve. We look at both in-sample and out-of-sample. Cubic splines offer the best in-sample fit, but out-of-sample results indicate that NS and DL curves might be a better choice for researchers interested in medium- and longer-dated securities. We also assess the usefulness of short-term interbank offered rates as proxy for the shortest part of the yield curve, finding no appreciable empirical benefit and lack of theoretical justification.
No takes yet. Share an insight, caveat, or question.
Piotr Bartkiewicz (2024) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: