Test of Stationarity
The above illustrates an informal test of stationarity. Both inflation(infla) and growth (y_growth) appear to have a slow decay over time and so they do not seem to have conspicous trend and hence demonstrate some stationarity status. The informal test however would not give a clear picture on stationarity status of variables.
Levin & Chu Panel Unit Root test results
The stationarity status of the variables y_growth, Yp, Pr, Gc Opn, Gf and Infla is observed using the Levin, Lin & Chu (LLC) panel unit root test. The LLC is based on the assumption that there is a common persistence of parameters across cross-sections.
The test equation used for the stationarity test for each of the variables is based on the graphical illustration as to whether each variable has an intercept or trend.
Among all the variables used for this study, only two are stationary in their level form. However, the variables tagged institutionalized democracy (Dem) is left out since it takes the form of a dummy representation in the analysis.
The growth variable (y_growth) is just stationary at level with probability value 0.000 which is less than the conventional 0.05 level. As expected, the inflation variable (infla) is also found with no unit root at its level (0.000). On this basis, both the growth and inflation variables are integrated of order 0.
Other variables such as Per capita GDP (Yp), private investment (Pr), government final consumption (Gc), Openness (Opn) and gross fixed capital formation (Gf) are stationary only after taking their first differences, hence they are integrated of order 1. The overall results of the stationarity test follow from the fact that most economic variables appear to be integrated in their first differences.
Kao-residual co integration
Pedroni residual cointegration
The above tables discuss the long run relationship among the variables used. First, the cointegration test is performed for three different equations. These are the growth (Y_growth), the economic development (Per capita GDP) and the government final consumption (Gc) equations. The three variants of panel cointegration tests carried out are the Johansen Fisher Panel cointegration test, the Kao Residual cointegration test and the Pedroni Residual cointegration test. These become necessary to ascertain the robustness of the estimation results.
In the test for the long run relationship among the variables using the Johansen Fisher Panel cointegration test, a linear deterministic trend is assumed with lag interval of 11 in first differences.
For the growth equation where growth is endogenously defined and determined by the private investment, government final consumption, state of democracy, openness, government fixed capital formation; there is evidence of a long run relationship among these variables.
The Johansen cointegration test has demonstrated five cointegrating vectors for both the Trace test statistics and the Max-Eigen value statistics; all with probability values 0.000. The number of cointegrating vectors displayed is a clear indication of the stability of the system since the greater the number of the cointegrating vectors, the more stable the system is.
The Kao Residual test gives similar results of a long run relationship as demonstrated by the t-statistics of −8.122. In the same way, the Pedroni cointegation test shows that the variables are cointegrated especially for the Panel PP statistics (−5.243), Panel ADF statistics (−4.43), Group PP statistics (−5.047) and the Group ADF statistics (−3.697).
The long run relationship among the variables in the growth equation follows the interaction between the determinants and growth. It further demonstrates that these variables seem to be fundamental to growth in the ECOWAS region.
The Economic development equation shows similar trend. In this case too using all the three variants of cointegration tests, there exists a long run relationship among the variables in the economic development equation including the endogenous variable itself. Exactly five cointegrating vectors too were displayed for both Trace and Max-Eigen value statistics. This also shows that the model operates under a stable system. The Kao statistics also shows evidence of a long run relationship among the variables in question based on its ADF t-statistics of −8.239. The Pedroni cointegration test gives similar results of cointegration among the variables as observed from the Panel PP statistics (−7.819), Panel ADF statistics (−2.817) Group PP statistics (−7.276) and Group ADF statistics (−2.231).
All the three variants also demonstrate a long run relationship between the government final consumption and its determinants.