3.2.4. Granger Causality

One variable, *x*, is defined as having Granger causality on variable *y* if the given previous information of *y*, as well as the past values of *x*, enable to forecast the current value of *y*. This is the fundamental concept of Granger causality of these variables. Therefore, the null hypothesis is to test that the joint coefficients equal zero, including the lagged values of *x* in the regression between *y*, the lagged values of *y* and *x*. Therefore, Granger (1969) employed the Wald statistics for the hypothesis. Especially, the procedure for the Granger causality test is to store all values in VAR (or SVAR) regression first. Afterwards, they calculate and report small sample F statistics or large sample *χ*2 statistics for the null hypotheses.
