We study the economic sources of stock‐bond return comovements and their time variation using a dynamic factor model. We identify the economic factors employing a semistructural regime-switching model for state variables such as interest rates, inflation, the output gap, and cash flow growth. We also view risk aversion, uncertainty about inflation and output, and liquidity proxies as additional potential factors. We find that macroeconomic fundamentals contribute little to explaining stock and bond return correlations but that other factors, especially liquidity proxies, play a more important role. The macro factors are still important in fitting bond return volatility, whereas the “variance premium” is critical in explaining stock return volatility. However, the factor model primarily fails in fitting covariances. (JEL G11, G12, G14, E43, E44) Stock and bond returns in the United States display an average correlation of about 19% during the post-1968 period. Shiller and Beltratti (1992) underestimate the empirical correlation using a present value with constant discount rates, whereas Bekaert, Engstrom, and Grenadier (2005) overestimate it in a consumption-based asset pricing model with stochastic risk aversion. Yet, these models generate realistically positive correlations using economic state variables.