Dividend yields are a determinant of asset prices, but changes in dividend growth impact both dividend yields and discount rates. As a result, dividend growth is typically treated as a known constant in most of the literature.
In this paper, we develop a dynamic approach to forecasting dividend growth using Bayesian filtering techniques, which improves markedly on standard linear methods. The resulting growth-adjusted dividend yield improves out-of-sample return predictions by several orders of magnitude. These results show that dynamic cash flow modeling can significantly improve the performance of expected return models.