Electricity Consumption and Economic Growth in Turkey: Cointegration, Linear and Nonlinear Granger Causality

Abstract

The purpose of this study is to examine causal relationships between the electricity consumption and economic growth in Turkey for the period 1967–2007. To accomplish this purpose, we utilize three analytical tools in time series econometrics: The bounds testing cointegration approach, the linear Granger causality test and the nonlinear Granger causality test. We depart from the existing literature on the energy consumption and economic growth nexus in Turkey by testing nonlinearity in the series and carrying out the nonlinear Granger causality test. The cointegration analysis shows that the electricity consumption and economic growth are cointegrated in the long-run. The linear Granger causality test based on the error correction model shows the bi-directional Granger causality in both the short- and the long-run between the electricity consumption and economic growth in Turkey. After filtering out the linear properties of the series with cointegration analysis, we carry out the Brock, Dechert, and Scheinkman nonlinearity test and find evidence of nonlinearity in the series. Therefore, the inferences from the nonlinear Granger causality test are more appropriate than the linear Granger causality analysis for Turkey. The nonlinear causality analysis supports the neutrality hypothesis, implying that electricity conservation policy will not damage the growth in the Turkish economy.

Keywords:

• cointegration
• economic growth
• electricity consumption
• linear and nonlinear causality
• Turkey

Notes

¹See Ozturk (2010) for a survey of the literature.

²According to Belloumi (2009), a small share of energy costs in production may result in no causal feedbacks between energy consumption and economic growth.

³Since the reform of electricity sector in Turkey is well documented in Balat (2009) we do not outline it here in order to save space.

⁴The choice of this period is based on the data availability.

*** and ** denote statistical significance at the 1% and 5% level, respectively.