Abstract
This article investigates the out-of-sample predictability of bond excess returns. I assess the statistical and economic significance of forecasts generated by empirical models based on forward interest rates, macroeconomic variables and risks in macroeconomic outcomes. Results suggest that macroeconomic variables, risks in macroeconomic outcomes as well as combinations of different predictors outperform a constant model of no-predictability. These results are confirmed when using macroeconomic data available in real-time, suggesting that the predictability of bond returns is not driven by data revisions.