Adaptive Dynamic Nelson-Siegel Term Structure Model with Applications

Abstract

We propose an Adaptive Dynamic Nelson-Siegel (ADNS) model to adaptively forecast the yield curve. The model has a simple yet flexible structure and can be safely applied to both stationary and nonstationary situations with different sources of change. For the 3-to 12-months ahead out-of-sample forecasts of the US yield curve from 1998:1 to 2010:9, the ADNS model dominates both the dynamic Nelson-Siegel (DNS) and random walk models, reducing the forecast error measurements by between 30 and 60 percent. The locally estimated coefficients and the identified stable subsamples over time align with policy changes and the timing of the recent financial crisis.

A local vector autoregressive framework and its applications to multivariate time series monitoring and forecasting

“…In fact, as a specific case of this general setting, Chen and Niu (2013) show that restricting the state dynamics to an AR(1) model for each NS factor greatly improves forecast accuracy, and that the resulting performance beats the alternative rolling or recursive forecast uniformly. The reason may be due to the off-diagonal elements in the VAR autoregressive coefficient matrix being typically sparse and close to zeros, which does not contribute much to the improvement of forecast accuracy, but deteriorates the information efficiency.…”

“…The Nelson-Siegel (NS) model is parameterized according to Diebold and Li (2006), and the data set is a fifteen yield series as used in Chen and Niu (2013). We denote the underlying parameters as Θ 0 = (c 0 , A 0 , Σ 0 ) and keep them constant throughout the whole sample.…”

“…Chen, Härdle and Pigorsch (2010) propose the LAR detection and estimation methodology and successfully apply it to forecasting realized volatility in a financial time series. Chen and Niu (2013) extend the method to an LARX model and apply it to yield curve modeling and forecasting, where under a traditional dynamic Nelson-Siegel factor model Diebold and Li (2006) each state yield factor is modeled as an LARX with inflation as the exogenous variable. Chen and Niu (2013) demonstrate that the data-driven LARX model brings clear advantages in forecasting the whole yield curve, when compared to a wide spectrum of traditional yield curve models with rolling or recursive estimation strategies using predetermined estimation windows.…”

“…Chen and Niu (2013) extend the method to an LARX model and apply it to yield curve modeling and forecasting, where under a traditional dynamic Nelson-Siegel factor model Diebold and Li (2006) each state yield factor is modeled as an LARX with inflation as the exogenous variable. Chen and Niu (2013) demonstrate that the data-driven LARX model brings clear advantages in forecasting the whole yield curve, when compared to a wide spectrum of traditional yield curve models with rolling or recursive estimation strategies using predetermined estimation windows. Moreover, the detected local intervals and the resultant parameter evolutions estimated in real-time are of great help for monitoring the yield curve dynamics, understanding its changing relationship with the inflation factor, and providing evidence on policy changes and on changing interest rate behavior during the recent financial crisis.…”

LCARE - Localizing Conditional Autoregressive Expectiles

“…The main advantage of the approach is the achievement of a balance between modelling bias and parameter variability. This approach has been successfully applied in many research areas: Čížek et al (2009) analyse the GARCH(1, 1) models, Chen et al (2010) explore it to forecast realised volatilities, Chen and Niu (2014) predict the interest rate term structure, whereas Härdle et al (2015) utilise it successfully in high frequency time series modelling and forecasting.…”