Testing Granger Causality under Dynamic Covariance

The paper considers the Granger causality tests based on the hetersokedasticity-consistent covariance matrix of White (1980). The Monte Carlo experiments show that the conventional test over-rejects the null hypothesis of non-causality, while the proposed test works satisfactory. The proposed test also has enough power. Using the data for the Standard and Poor's 500 Composite Index, the paper examines the causalities among the return, volatility and trading volume.departmental bulletin pape