We develop a volatility decomposition derived from flexible and robust local projections to quantify the relative contributions of expected discount rates and cash flows to the variation of dividend yields. Local projections enable the incorporation of large information sets, the use of monthly data along with annual data, and to consider time variation in the volatility decomposition. While the variation of expected discount rates remains the dominant contributor to market volatility, we find that the contribution of expected cash flows is non-negligible when moving beyond the standard model with the dividend yield as the single state variable.