Filtering Out Expected Dividends and Expected Returns

This paper applies a state space approach to the analysis of stock return predictability. It acknowledges that expected returns and expected dividends are unobservable and uses the Kalman filter to extract them from the observed history of realized dividends and returns. The suggested approach explicitly takes into account the time variation in expected dividend growth rates and exploits the present value relation. The obtained predictors for future returns are robust to structural breaks in the means of expected dividends and returns and more efficient than the dividend price ratio. The likelihood ratio test reliably rejects the hypothesis of constant expected returns.