Accordingly, the non-stationary I(1) lnOFDI and lnGDP, which are to be included in the
VAR specification for Granger non-causality test, have to be differenced once to ensure
all variables are stationary within the VAR system. Thus, the prospective causality
between lnOFDI and lnGDP can be written in a bi-variate VAR as follows: