Abstract
This paper documents novel evidence that the ability of the dividend-price ratio (DP) to predict market excess returns is contingent on the level of war. We find that DP predicts future returns primarily in high-war periods, whereas this ratio holds no predictive power in low-war periods. This pattern holds across short and long horizons, survives both in-sample and out-of-sample tests, is supported by both statistical and economic metrics, and extends to other standard valuation ratios. To rationalize these findings, we develop a war-dependent present-value framework that links war regimes to DP persistence and return predictability via a mispricing correction channel. Our model shows that in high-war periods, mispricing corrects rapidly, and DP becomes less persistent and exhibits strong predictive power for stock returns. In low-war periods, mispricing persists and DP becomes more persistent, and its predictive power is substantially weaker.