Regime dependent causality relationship between energy consumption and GDP growth: evidence from OECD countries

ABSTRACT

This study empirically investigates the energy consumption-GDP growth nexus for the period from 1971 to 2016 for 26 OECD countries. The prevailing studies in the literature use limited econometric methodologies, which may wrongly model the underlying relationship and lead to misleading policy conclusions. Our study utilizes the newest econometric methods to reveal the nonlinear relationships in the long-run. Furthermore, to capture the asymmetric behaviour of regime changes, four residual-based nonlinear cointegration tests are implemented. Finally, a two-regime TAR type of panel threshold VECM model (PTVAR) is estimated for testing the presence of nonlinear short- and long-run causality. Our findings indicate a state-dependent causality between energy consumption and GDP growth.

KEYWORDS:

• Residual and ECM-based panel nonlinear cointegration tests
• energy Consumption and GDP Growth
• panel threshold causality test

JEL CLASSIFICATION:

• G1
• C12
• C23
• Q43

Notes

² After the U.S. found the Shale gas, the relationship between energy consumption and the growth again changed after 2015–16. So our study didn’t include this dynamic into the analysis. However, for further study the new era must be analysed after sufficient data become available. The OECD group is selected for this study. The OECD group is the well-behaved group that the economic stability is high with respect to other groups of countries. Unfortunately, there are lots of studies conducted in this group and still the results are mixed and scant. Therefore, the results must be checked with more advance econometric methodology.

³ Australia, Austria, Belgium, Canada, Chile, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Israel, Italy, Japan, Korea, Luxembourg, Mexico, the Netherlands, Norway, Portugal, Spain, Sweden, Turkey, the United Kingdom, and the United States.

⁴ See Westerlund (2007) page 722–724

⁵ To tackle deterministic components in LR cointegration relation, they use the de-meaned and de-trended data.

⁶ For details, see Omay, Emirmahmutoglu, and Denaux (2017) page 3.

⁷ For further readings about the threshold estimation and the misspecification tests, see Omay (2014) and Omay, Emirmahmutoglu, and Denaux (2017).