Empirical Analysis and Forecasting of Multiple Yield Curves
Posted: 10 Jan 2019 Last revised: 1 Sep 2020
Date Written: January 8, 2019
In this paper we develop new dynamic factor models to forecast multiple yield curves. Our methodology is based on a thorough empirical study of daily tenor-dependent term structures over the time period 2005-2017 which reveals important cross-tenor dependencies of yields. The suggested forecasting approach accounts for these dependencies and thus targets the new features of post-crisis interest rate markets. Our method generates extremely precise predictions of future yield curves for various forecasting horizons. In particular, it clearly outperforms existing single-curve forecasting methods which naturally omit any connections between rates of different tenor structures.
Keywords: multiple term structures, principal component analysis, dynamic factor model, Nelson-Siegel curve, forecasting of yield curves
JEL Classification: G1, E4, C5, C3
Suggested Citation: Suggested Citation