Abstract
In this paper we investigate whether information in credit spreads helps improve the forecasts of government bond yields. To do this, we propose and estimate a joint dynamic Nelson-Siegel (DNS) model of the U.S. Treasury yield curve and the credit spread curve. The model accounts for the possibility of regime changes in yield curve dynamics and incorporates a zero lower bound constraint on yields. We show that our joint model produces more accurate out-of-sample density forecasts of bond yields than does the yield-only DNS model. In addition, we demonstrate that incorporating regime changes and a zero lower bound constraint is essential for forecast improvements.
Original language | English |
---|---|
Pages (from-to) | 39-61 |
Number of pages | 23 |
Journal | Journal of International Money and Finance |
Volume | 64 |
DOIs | |
Publication status | Published - 2016 Jun 1 |
Keywords
- Bayesian MCMC estimation
- Density prediction
- Dynamic Nelson-Siegel
- Predictive likelihood
ASJC Scopus subject areas
- Finance
- Economics and Econometrics