The predictability of asset returns in the BRICS countries: a nonparametric approach

One of the earliest and most enduring questions of financial econometrics is the predictability of financial asset prices. In this article, stock market data from Brazil, Russia, India, China and South Africa are used to assess the out-of-sample performance of the ARMA(1,1)-GARCH(1,1) and Non-parametric kernel (Epanechnikov) regression models. The results reveal that the non-parametric kernel regression model outperforms its parametric rival based on the predicted mean square error (PMSE), Diebold-Mariano criterion, Mean-Absolute Deviation (MAD) and Variance statistics. These results confirm those found previously by other researchers whereby non-parametric forecasting models outperform parametric models in the short-term forecasting horizon.