**3. Methodology**

The current section will provide a summary of the methodologies used in the series of papers the current review focuses on. Those methodologies are varied depending on the different characteristics of the data and the diagnosed problems. The section is further divided into three sub-sections on unit root testing, cointegration and causality, respectively. The section serves an informative purpose because the current study is a review of previous studies and not a new empirical one. Thus, this section does not describe the employed methodologies from scratch as this can be done in the relevant original papers. These papers are: Menegaki and Tugcu [33], Menegaki et al. [34], Menegaki and Tugcu [35], Menegaki and Tiwari [31], Menegaki and Tugcu [17], Menegaki and Tugcu [30].

#### *3.1. Unit Roots Testing*

Testing for unit roots is the first stage in every study in the energy-growth nexus. First, the study by Menegaki and Tugcu [35] in page29, uses the cross-sectionally augmented IPS test for the investigation of unit roots in cross-sectionally dependent data. Previously, in page 29–31 they have used the Pesaran CD test [36], Pesaran scaled LM test [36] and Baltagi et al. [37] bias-corrected scaled LM test. The presence of cross-sectional dependence was confirmed in three out of four statistics. Second, the study by Menegaki et al. [34] in page 1261 has employed ADF Fisher, PP Fisher (Maddala and Wu, [38] and Choi, [39]) and CIPS (Zt-bar) test (Pesaran, [40]). Third, the study by Menegaki and Tugcu [35] has employed the Pesaran IPS and GIPS (Pesaran et al. [40]) test in page 896 in constant and constant and trend versions. Fourth, the study by Menegaki and Tiwari [31] in pages 499–500 employs a battery of unit root tests such as the Levin et al. [41], the Im et al. [42], the augmented Dickey-Fuller test [43] the Phillips-Perron test [44] and the Breitung *t*-test [45] Hadri Z-test [46] and Heteroskedastic consistent Z test. Fifth, the study by Menegaki and Tugcu [30] in page 81 tests panel unit roots with Im et al. [41] and Choi [47]. None of

the variables was stationary at levels, so they took first differences. Sixth, the study by Menegaki and Tugcu [33] in pg 156 has also used the augmented IPS (Pesaran, [48]), after acknowledging the existence of cross-sectional dependence (Pesaran, [40]).
