Stock Return Predictability and Dividend-Price Ratio: A Nonlinear Approach

This paper examines the forecasting ability of the dividend-price ratio for international stock market returns. Hitherto, existing research has only considered this issue in-sample and in a linear framework. Hence, this paper provides the first systematic study of non-linear forecasting within the present value model context. Using an asymmetric variant of the popular ESTR model we demonstrate the superior forecasting ability for the G7 markets over both a linear and symmetric ESTR model on the basis of a variety of forecast performance test. In particular, the asymmetric-ESTR model provides improved mean forecast error metrics that are largely significant on the basis of forecast equality tests. Furthermore, in a trading rule exercise this models provides superior trading returns. As a final exercise we compare the forecasting performance of the individual models to those obtained through forecast combination. These results support the individual models on the basis of forecast error test but suggest the combination strategy may be more profitable. These results are of importance not only for model builders but also market participants looking to undertake active investment strategies.