Modeling and Forecasting the Yield Curve Using Time-varying Parameters

We use the Nelson-Siegel framework to fit and model the dynamics of the yield curve. We compare models that allow for a time-varying decay parameter to the widely used two-step-approach that was introduced by Diebold and Li (2006). Next to the different models to fit the yield curve, we introduce a novice method, recently used in forecasting realized volatility, to model the latent factors over time. This method includes the standard errors of the factor coefficients that are obtained in the first step of the process. We find that there is a clear trade-off between the in-sample fit and the out-of-sample forecasting power. Robust factors prove to be essential for the out-of-sample forecasting power.