Note(s): Asterisks denotes significance level at 10% (*), 5% (**) and 1% (***). The numbers in parenthesis indicate t-statistics. In this study, we used the two-step system GMM as our baseline model. We conduct the Durbin-Wu-Hausman test for the endogeneity test of all regressors as a group. Endogeneity testing via Durbin-Wu-Hausman implies a rejection of the null hypothesis that the regressors are exogeneous; hence the system GMM necessarily be used. In addition, Hansen’s J-statistic results show that we cannot reject the null hypothesis that the instrumental variables used in the system GMM model are valid, which supports the validity of over-identifying restrictions and the arguments for using the system GMM. The F-statistic value is significant, indicating that the model has predictive power. AR(1) shows the presence of first-order autocorrelation of residuals, whereas AR(2) indicates that there is no second-order autocorrelation. The p-values of both Chow and Breusch-Pagan LM tests indicate that either fixed-effect or random-effect regression procedures should be used, based on the results. The Hausman fixed, or random (F/R) test findings confirm that the fixed effect is the optimal approach. Year dummies and country dummies are unreported