Resumen
In this paper, we examined and compared the forecast performances of the dynamic Nelson?Siegel (DNS), dynamic Nelson?Siegel?Svensson (DNSS), and arbitrage-free Nelson?Siegel (AFNS) models after the financial crisis period. The best model for the forecast performance is the DNSS model in the middle and long periods. The AFNS is inferior to the DNS model for long-period forecasting. In U.S. bond markets, AFNS is shown to be superior to DNS in the U.S. However, for Japanese data, there is no evidence that the AFNS is superior to the DNS model in the long forecast horizon.