Nonlinearities, cyclical behaviour and predictability in stock markets: International evidence

In this paper we employ the STAR (smooth transition autoregressive) model to investigate potential nonlinearities and cyclical behaviour in the stock prices of seven major industrial countries (the G-7). Tests reject linearity for all stock markets. The estimated nonlinear models suggest that stock price growth rates are characterised by asymmetric cycles in most countries, with the speed of transition between the expansion and contraction regimes being relatively slow for all countries. The implied transition probabilities detect satisfactorily the main contraction regimes in stock markets. STAR-based noncausality tests indicate only a small number of interactions between the stock markets. Our evidence on out-of-sample forecasting suggests that forecast gains can be made by exploiting the nonlinear structure of STAR models.