Non-linear predictability in stock and bond returns: When and where is it exploitable?

Abstract

We systematically examine the comparative predictive performance of a number of linear and non-linear models for stock and bond returns in the G7 countries. Besides Markov switching, threshold autoregressive (TAR), and smooth transition autoregressive (STAR) regime switching models, we also estimate univariate models in which conditional heteroskedasticity is captured by GARCH and in which predicted volatilities appear in the conditional mean function. We find that capturing nonlinear effects may be key to improving forecasting. In contrast to other G7 countries, US and UK asset return data are "special," requiring that non-linear dynamics be modeled, especially when using a Markov switching framework. The results appear to be remarkably stable over time, robust to changes in the loss function used in statistical evaluations as well as to the methodology employed to perform pair-wise comparisons.