The paper by John Y. Campbell and Robert J. Shiller explores the relationship between stock prices, earnings, and expected dividends. Using historical data from the U.S. stock market (1871–1986), they find that long historical averages of real earnings help forecast future real dividends. The authors develop a vector-autoregressive (VAR) model to forecast the present value of future dividends, which is roughly a weighted average of moving-average earnings and current real prices, with the earnings measure carrying more weight. They argue that this finding has implications for the present-value model of stock prices and recent results showing that long-horizon stock returns are highly forecastable. The authors introduce earnings data into a VAR framework, accounting for measurement errors and allowing earnings to enter the model only if they are useful for forecasting. They use this approach to answer two questions: what components of stock returns can be predicted given the information used in the VAR system, and what components can be accounted for ex post by news about future dividends. The results show that stock returns and dividend-price ratios are too volatile to be fully explained by news about future dividends, and that this excess volatility is closely related to the predictability of multiperiod returns. The paper also discusses the implications of these findings for the constant-expected-real-return model of stock prices. The authors find that the log dividend-price ratio has only a weak relationship with its theoretical counterpart, contradicting the model. They conclude that stock prices and returns are much too volatile to accord with a simple present-value model, yet annual returns do carry some information and are correlated with what they should be given the model.The paper by John Y. Campbell and Robert J. Shiller explores the relationship between stock prices, earnings, and expected dividends. Using historical data from the U.S. stock market (1871–1986), they find that long historical averages of real earnings help forecast future real dividends. The authors develop a vector-autoregressive (VAR) model to forecast the present value of future dividends, which is roughly a weighted average of moving-average earnings and current real prices, with the earnings measure carrying more weight. They argue that this finding has implications for the present-value model of stock prices and recent results showing that long-horizon stock returns are highly forecastable. The authors introduce earnings data into a VAR framework, accounting for measurement errors and allowing earnings to enter the model only if they are useful for forecasting. They use this approach to answer two questions: what components of stock returns can be predicted given the information used in the VAR system, and what components can be accounted for ex post by news about future dividends. The results show that stock returns and dividend-price ratios are too volatile to be fully explained by news about future dividends, and that this excess volatility is closely related to the predictability of multiperiod returns. The paper also discusses the implications of these findings for the constant-expected-real-return model of stock prices. The authors find that the log dividend-price ratio has only a weak relationship with its theoretical counterpart, contradicting the model. They conclude that stock prices and returns are much too volatile to accord with a simple present-value model, yet annual returns do carry some information and are correlated with what they should be given the model.