Performing Granger non-causality test in a VAR (vector autoregressive) framework
assumes the endogenous variables entering into the VAR system to be stationary (i.e.
I(0)). Conversely, estimating a VAR system using non-stationary variables (i.e. integrated
of order one, I(1)) may result bias in inference. Therefore, it is necessary to examine the
time series properties (i.e. the degree of integration, I(d)) of real OFDI and real GDP in
this study. Two different unit root tests are applied in order to assume consistency,
namely the Phillips-Perron (PP) unit root test (Phillips and Perron, 1988) and
Kwiatkowski-Phillips-Schmidt-Shin (KPSS) stationary test (Kwiatkowski et al., 1992).
The former assumes each series to be a unit root under the null and allows for
heterogeneous residuals of a unit root process, while the latter assumes the null is stationary.
Table 1 presents the results of PP and KPSS tests for both data in levels and first
differences (as symbolized by ¨). Both PP and KPSS test statistics show that the
variables lnOFDI and lnGDP are integrated of order one (I(1)), concluding that both
series have a unit root. Hence, both variables are first differences once and both tests
confirm they are stationary I(0)