Mixing information from macroeconomic fundamentals and yields helps to understand bond return predictability. We derive the term-structure of interest rates from a structural setting with CRRA preferences and flexible dynamics of macroeconomic growth rates that puts formal restrictions between factor dynamics and risk premia. By letting conditional volatilities of the growth rates fluctuate through time, we generate a quadratic term-structure model along with time-variation in bond risk-premia that is determined endogenously. Estimating the parameters from yield curve and the macroeconomic fundamentals jointly, we conclude that by effectively using prior information to mix bond yields and macroeconomic variables, we can predict bond returns better than from either matching time-series of yields or macroeconomic fundamentals. The posterior distribution of R-squares obtained from this partial equilibrium setting can be as high as 8-10% on bond returns of maturity 2-5 years. Furthermore, bond risk-premia is small and largely negative in the post-Volcker regime.