I study the sufficiency of macroeconomic information to explain the time-variation in second moments of stock and bond returns, with a particular attention to stock-bond correlations. I propose an external habit model supplemented with realistic non-Gaussian fundamentals estimated solely from macroeconomic data. Intertemporal smoothing and precautionary savings effects – driven by consumption shocks – combine with a time-varying covariance between consumption and inflation to generate large positive and negative stock-bond return correlations. Macroeconomic shocks are most important in explaining second moments of stock and bond returns from the late 1970’s to mid-1990’s and during the Great Recession.

We use non-Gaussian features in U.S. macroeconomic data to identify aggregate supply and demand shocks while imposing minimal economic assumptions. Macro risks represent the variables that govern the time-varying variance, skewness, and higher-order moments of these two shocks, with ”good” (”bad”) variance associated with positive (negative) skewness. We document that macro risks significantly contribute to the variation of yields and risk premiums for nominal bonds. While overall bond risk premiums are countercyclical, an increase in aggregate demand variance significantly lowers risk premiums. Macro risks also significantly predict future realized bond return variances.