Pesaran et al. (2001) have proved that the distribution of this F-statistic is non-standard
irrespective of whether the regressors are I(0) or I(1), and have tabulated the appropriate
critical values. Depending on the number of regressors and on whether an intercept
and/or a time trend is included in the equation, a pair of critical values is provided, which
constitute an upper and a lower bound respectively. If the F-statistic is greater than the
upper bound, the null hypothesis is clearly rejected and a long-run relationship exists
among the test variables. If the F-statistic is smaller than the lower bound, then the null
cannot be rejected and estimation can continue assuming no long-run relationship. If the
statistic falls between the two bounds, then the result is inconclusive; it is only at this
stage that the analyst may need to conduct unit root tests in order to proceed (Pesaran and
Pesaran, 1997).
The long-run relationship test is equivalent to the cointegration test. If such a relationship
is found in Eq (1) according to the bounds test described above, this would imply longrun
causality from FDI to GDP. Short-run causality in the same direction can be tested
through a standard Wald or F-test for the joint significance of coefficients β2. To test
causality from GDP to FDI, one has to formulate an equation similar to Eq (1) but using
FDI as the dependent variable and GDP as the exogenous one and employ the same tests
as outlined above