Empirical Differences between the Overnight and Day Trading Hour Returns : Evidence from the Chinese Commodity Futures

In this study, commodity indices are constructed using principal components analysis to represent the market returns for day and night trading in the Chinese commodity futures market. Then VAR models are employed to predict the commodity indices' returns and squared returns. The symmetric VAR model failed to model the market returns since the asymmetric effects of positive and negative returns are not taken into account. By allowing asymmetric behavior among positive and negative variables, asymmetric VAR model is utilized to trace the leading effect of overnight returns to day trading returns. However, the symmetric VAR model outperforms the asymmetric model when evaluating the predictive power of squared returns during night trading hours