Forecasting the yield curve with the arbitrage-free dynamic Nelson-Siegel model : Brazilian evidence

We assess the extent to which the imposition of a no-arbitrage restriction on the dynamic Nelson-Siegel model helps obtaining more accurate forecasts of the term structure. For that purpose, we provide an empirical application based on a large panel of Brazilian interest rate future contracts and test for differences in forecasting performance among alternative benchmark specifications including the random walk, vector autoregressions, and the dynamic Nelson-Siegel. We show empirically that the arbitrage-free Nelson-Siegel model is able to outperform all other benchmark models when longer forecasting horizons are taken into account.