Proposal of a new panel unit root test augmented with residuals: RALS-CIPS test

Stationarity testing of macroeconomic variables was previously analyzed by time series graph, correlogram graph or autocorrelation functions. With the developing literature, the stationarity tests of macroeconomic variables are now carried out by unit root tests. For this reason, whether it is time series or panel data structure, unit root testing is a very important issue. Because econometric analysis with non-stationary variables can be inconvenient and have false and misleading results. The obtained t statistics give meaningful results and the coefficient of determination R^2 is quite high. In such cases, a spurious regression problem may arise. In other words, although the parameters seem to be related in a regression relationship to be established between two non-stationary series, it is meaningless in terms of the theorem or theory examined. Even working in large samples cannot prevent this spurious regression problem. One of the main assumptions in the regression analysis, which forms the basis of econometrics, is the normal distribution of residues. However, it is sometimes can't be controlled whether this assumption has been violated or not. In the case of non-normal distribution, analyzing under the assumption of normal distribution can cause biased and misleading results. For this reason, a new test process has been proposed, using the knowledge that residues in the CIPS test structure proposed by Pesaran (2007) do not distribute normally. A new test called RALS-CIPS has been developed by adding the second and third moments of the error terms to the CADF regression that constitutes the CIPS testing process. The test proposed in this study is a test that is powerful when error terms are not normally distributed. An empirical application study has also been carried out with this new proposed RALS-CIPS test. For this purpose, the existence of unit root in the gross domestic product per capita of 15 EU countries was investigated.