Stock return predictability: Evidence from a structural model

We derive a linear model that can be used to explain variations in stock returns. Our derivation is based on the dividend discount theory. The model incorporates macroeconomic variables through the Taylor rule, which makes it also relevant for policy-makers. One advantage of this model is that its parameters are transparent, thereby permitting an examination of the sources of the well documented instability of parameters in return prediction models. We estimate the model for six advanced and five emerging market economies and find that its ability to explain variations in stock returns is a significant improvement on existing models in the literature. Out-of-sample forecast evaluation shows that the model consistently beats the historical average benchmark and it beats the autoregressive benchmark at longer horizons. Further tests reveal that our model forecasts are better than those derived from the simple dividend yield model at longer horizons.