I am grateful to James Morley for his valuable discussions and suggestions. Also I would like to thank Bakhodir

Many studies find that government bond yield spreads predict real economic activity and, to lesser extent, inflation. Most of these studies define the yield spread exogenously as the difference between two yields of specific maturities to predict output and inflation. In this paper, I propose a different approach, which allows using information contained in the entire term structure of US Treasury yields to predict US real GDP growth and inflation. In particular, I modify the Diebold-Li (2006) dynamic yield curve model, which is based on the Nelson-Seigel (1987) three latent factor framework, to model the entire yield curve, real GDP growth, and inflation jointly. I find that this yield curve-macro model produces better out-of-sample real GDP forecasts than those produced by the simple yield spread model. Using the dynamic yield curve model provides a more accurate depiction of the predictive power of the yield curve and uses information contained in the entire yield curve, while the yield spread forecasting model uses only information in the difference between two yields of arbitrarily chosen maturities and also possibly overfits in sample. With regard to inflation, I do not find robust improvement in forecasts based on the dynamic yield curve model over the yield spread model and there is evidence of a weakening relationship between the yield curve and inflation in post-1982 period