Energy consumption and economic growth in Ethiopia: A dynamic causal linkage

Abstract

In this study, we have explored the causal relationship between energy consumption and economic growth in Ethiopia, during the period from 1971 to 2013. We have employed a multivariate Granger-causality framework that incorporates financial development, investment and trade openness as intermittent variables – in an effort to address the omission-of-variable bias. Based on the newly developed ARDL bounds testing approach to co-integration and the error-correction model-based causality model, our results show that in Ethiopia, there is a distinct unidirectional Granger-causality from economic growth to energy consumption. These results apply, irrespective of whether the estimation is done in the short run or in the long run. We recommend that policy makers in Ethiopia should consider expanding their energy-mix options, in order to cope with the future demand arising from the real sector growth.

Keywords

Ethiopia energy consumption economic growth Granger-causality

Introduction

The relationship between energy consumption and economic growth has been examined extensively in many studies. Various studies on the causal relationship have focused on different countries, different time periods, different methodologies and different energy-consumption variables.^1–6 Recently, attention has been drawn to develop cleaner alternative fuels from renewable sources, in order to reduce the harmful emission to air and to decrease the need for fossil fuel, all in an effort to boost economic growth and conserve the environment.^7,8

Although the studies on the causality between energy consumption and economic growth are numerous, the results have been largely mixed.^9–12 Four main hypotheses on the causal relationship between the two variables have emerged in the literature; and it is by these four hypotheses that the direction of causality between energy consumption and economic growth is established.¹³

The first hypothesis centres on the notion that energy consumption causes economic growth. This view is known as the energy-led growth hypothesis; and it implies that restrictions on the use of energy would adversely affect economic growth; while increases in energy consumption would contribute to economic growth.^14,15 The second hypothesis, known as the growth-led energy-consumption hypothesis, supports the view that economic growth causes energy consumption; and this is largely based on the pioneering work of Kraft and Kraft.¹⁶ This hypothesis suggests that the policy of conserving energy consumption could be implemented with little or no adverse effect on economic growth.^17–20

The third category focuses on a bidirectional causal relationship between energy consumption and economic growth. This is also known as the feedback hypothesis. This relationship implies that energy consumption and economic growth cause each other.²¹ The fourth hypothesis purports that no causality exists between energy consumption and economic growth. This is also known as the neutrality hypothesis; and it implies that policies in relation to conserving or expanding energy consumption have no effect on the economy.^22,23

Although research on this subject matter has been done on a number of countries in both developed and developing countries, to the best of our knowledge, very few studies have been done on Ethiopia. To date, energy consumption and economic growth nexus in Ethiopia has been discussed only in terms of impact and within a panel-study setting.^24–26

Against this backdrop, this paper, therefore, seeks to examine the energy consumption-growth nexus in Ethiopia, in order to determine the direction of causality between these two variables, and to inform policy. The rest of the paper is organised as follows: the upcoming section gives an overview of the trends in energy consumption and economic growth in Ethiopia. Subsequent sections review the literature and discuss the methodology employed to test the causal relationship. Finally, the results of the study are provided and the conclusions are presented.

Trends in energy consumption and economic growth in Ethiopia: An overview

Ethiopia has various energy sources, including hydroelectric, geothermal, natural gas, coal, biomass, solar and wind energy. In 1992, Ethiopia’s total energy production was 21.70 million tonnes of oil equivalent (Mtoe); and the country’s electricity consumption was 1.12 terawatt hours (TWh). These statistics increased to 29.50 Mtoe and 1.84 TWh in 2002, respectively.²⁷ By then, the country’s real GDP had increased from USD$5.76 billion in 1992 to USD$9.80 billion in 2002. Ten years later, Ethiopia’s energy production had increased by about 45% to 43.04 Mtoe in 2012; whilst its GDP had more than doubled to USD$24.66 billion at 2005 prices.²⁷

Ethiopia’s energy consumption has predominantly been based on traditional energy sources, such as fuel wood, charcoal and dung cakes.²⁸ More than 90% of the country’s total final energy consumption is accounted for by the use of traditional biomass fuels.²⁹ However, the total energy consumption per capita in Ethiopia is reported to be 0.40 tons of oil equivalent (toe), which is far below the average sub-Saharan energy consumption of about 0.80 toe. This makes Ethiopia’s energy consumption one of the lowest in the world.³⁰ Currently, the per capita consumption of electricity in Ethiopia remains relatively low at about 200 kWh per year, due to the heavy reliance on traditional biomass energy sources, such as wood fuels, crop residues and animal dung.³¹

The country’s energy policy has emphasised the need to transform from traditional to modern energy sources, in order to support the developmental requirements of the country. In addition, the use of traditional energy sources has caused the continued destruction of forestry resources for firewood, which has resulted in environmental problems, loss of productivity and an ecological imbalance.²⁸ With growing energy demand worldwide, the use of fossil fuels is becoming increasingly important.⁸ As such, the development of new technologies and strategies in the management of energy consumption is vital – not only in Ethiopia but all over the world, in order to meet the energy demands and to limit the production of carbon dioxide. Ethiopia’s energy policy, therefore, prioritises hydropower-resource development; and it also encourages an energy mix, where renewable energy, such as those of solar, wind and geothermal origin, can be developed.³⁰

In light of the growing demand for energy and increased fuel prices, together with severe air-pollution restrictions in the road transport sector, other recent studies, such as that of Iodice and Senatore³² have highlighted the importance of the fraction of emissions from road transport sector in total emissions – in a bid to draw attention to the need to move away from reliance on fossil fuels due to their carbon-emitting behaviour.

From the economic growth front, real GDP per capita growth fell sharply from 4.6% in 1983 to -13.9% in 1985. Between 1989 and 1990, although still in the negative region, GDP per capita growth improved from −3.6% to −0.74%.³³ From 2002, the GDP per capita growth rate remained in the positive region until 2012.³³ Figure 1 illustrates the trends in economic growth, as measured by GDP per capita and energy consumption, proxied by electricity consumption, in Ethiopia from 1983 to 2012.

Figure 1.

The trends in economic growth and energy consumption in Ethiopia (1983–2012).

Literature review

Theoretical literature review

Following the pioneer work of Kraft and Kraft,¹⁶ who explored the relationship between energy consumption and economic growth in the United States, the energy-growth nexus has become an interesting subject of study to many economists and researchers – as they try to establish the direction of causality between the two. The theoretical literature posits the existence of four categories under which the energy-growth causality flow can be placed. These are the growth hypothesis, the conservation hypothesis, the feedback hypothesis and the neutrality hypothesis.^34–36

The growth hypothesis postulates that the direction of causality flows from energy consumption to economic growth. Thus, this hypothesis places importance on energy consumption in driving the real economy, both directly as input into the production processes and indirectly as a capital and labour inputs complement.^6,37,38 The growth hypothesis renders the implementation of energy policy as influential in determining the levels of production and output. In the same vein, the energy conservation strategies and policies will have an adverse effect on economic growth.³⁹

The conservation hypothesis states that there is unidirectional causality from economic growth to energy consumption; and this is confirmed only if an increase in the economic development increases the demand for energy. This hypothesis asserts that it is the economic growth that leads to increased energy consumption, implying that restrictive energy policies and strategies can be implemented in an economy without adversarial effects on the national output. Thus, it is this conservation hypothesis that suggests that a country can be less dependent on energy inputs while implementing conservation-oriented policies without impeding economic growth.^6,40

The third category is the feedback hypothesis, which postulates the existence of a bidirectional causal relationship between energy consumption and economic growth. Thus, in this causal setting, energy conservation policies and strategies will result in deterioration of economic performance as much as decreased in economic growth will lead to reduced demand on energy. With feedback hypothesis, energy and economic policies will need to be implemented jointly.^18,35

Finally, there is the neutrality hypothesis that suggests that there is no causality between energy consumption and economic growth; and that these two variables are not correlated. Thus, according to this hypothesis, energy consumption and economic growth are not dependent on each other and changes in one variable do not cause changes in the other variable.^40,41

Empirical literature review

Following the pioneering works of Kraft and Kraft,¹⁶ a wide range of studies have examined the energy consumption and economic growth causal relationship. This research spans from country-specific case studies to multi-country studies. The results have, however, been mixed. Some of the studies found that the causality runs from energy consumption to economic growth; others found that it runs from economic growth to energy consumption; while the rest either found a bidirectional relationship or no causality at all, between the two variables.

Wolde-Rufael⁵ used Toda and Yamamoto’s Granger-causality test for China from 1952 to 1999; and they found that energy consumption Granger-causes economic growth. Bowden and Payne⁴² used the same approach in a study of the United States between 1949 and 2006; and they also confirmed the energy-led growth hypothesis. In a multi-country setting, the Granger-causality test was also used by Yu and Choi⁴¹ in five countries; and the study found that the energy-led growth thesis was validated in the Philippines. Pao and Tsai⁴³ used Granger-causality tests and found that energy consumption Granger-causes economic growth for the BRIC (Brazil, Russia, India and China) countries from 1965 to 2009; whilst Wolde-Rufael⁴⁴ found the same for Algeria, Benin and South Africa in a panel study of 17 African countries.

Apergis and Payne⁴⁵ used panel co-integration and error-correction modelling to examine the causal relationship between nuclear energy consumption and economic growth for a panel of sixteen countries for the period between 1980 and 2005. The results of the panel-vector error-correction model showed that there is a unidirectional causal flow from nuclear energy consumption to economic growth in the long run. Co-integration and vector error-correction modelling were used to confirm the energy consumption-led growth hypothesis in Shiu and Lam,⁴⁶ for China; Akinlo⁴⁷ for Nigeria; and Ciarreta and Zarraga⁴⁸ for a panel of 12 European Countries.

Odhiambo¹³ used the autoregressive distributed lag (ARDL) bounds-testing procedure for Tanzania, and confirmed the energy-led growth hypothesis. Al-Mulali and Sab²⁴ used panel co-integration for sub-Saharan African countries; and they confirmed the energy-led growth hypothesis. Iyke⁴⁹ examined the dynamic causal linkage between electricity consumption and economic growth in Nigeria within a trivariate VECM, for the period 1971–2011. The results show that there is a distinct causal flow from electricity consumption to economic growth both in the short run and in the long run.

Azam et al.⁵⁰ examined the energy-growth causality nexus in ASEAN-5 countries during the period from 1980 to 2012. Based on Granger-causality testing approach utilised, the results came out in support of the growth hypothesis, where energy consumption was found to cause economic growth. Fang and Chang⁵¹ also examined the causal relationship between energy consumption and economic development in the Asian Pacific countries for the period between 1970 and 2011. Using the bootstrap-panel Granger-causality test, the results showed that in India, Korea, Pakistan and Taiwan, energy consumption Granger-causes GDP. The growth hypothesis was also supported in separate studies by Tang et al. and Destek.⁵²

Other studies have, however, validated the growth-led energy consumption hypothesis that postulates that economic growth drives energy consumption. Kraft and Kraft¹⁶ used a Granger-causality approach and found a unidirectional causal relationship running from GNP to energy consumption for the United States for the period from 1947 to 1974. The same approach was used by Cheng and Lai⁵³ for Taiwan for the period from 1954 to 1993 and they found that economic growth Granger-causes energy consumption.

Hsiao’s version of the Granger-causality approach was also employed for Japan by Cheng⁵⁴ for the period from 1952 to 1995. The results also confirmed a unidirectional causal flow from economic growth to energy consumption. Zhang and Chen⁵⁵ also found the same results when using the Granger-causality methodology for China for the period from 1960 to 2007. Stern and Enflo⁵⁶ found a causal flow from output to energy consumption in Sweden – after also employing the Granger-causality test from 1850 to 2000.

Aqeel and Butt⁵⁷ employed co-integration techniques, together with Hsiao’s version of the Granger-causality method, for Pakistan from 1955 to 1996; and they confirmed the growth-led energy consumption hypothesis. Yu and Choi⁴¹ examined the causal relationship between energy consumption and economic growth in five countries, using the Granger-causality test, from 1950 to 1976; and they found the growth-led energy consumption hypothesis to hold in Korea. Erol and Yu⁵⁸ found the same relationship in Italy and Germany; whilst Lee⁵⁹ also concluded the same for France, Italy and Japan.

In Asia, Chiou-Wei et al.¹⁵ found a causal relationship from economic growth to energy consumption in the Philippines and Singapore, using the Granger-causality test in a multi-country setting. Ang⁶⁰ used the Johansen co-integration and VEC modelling technique on Malaysia from 1971 to 1999; and they confirmed the growth-led energy consumption hypothesis.

Ouedraogo⁶¹ also confirmed the growth-led energy consumption hypothesis for 15 countries in the Economic Community of Western African States (ECOWAS) in the short run, using panel co-integration methods, based on the data from 1980 to 2008. Al-Iriani⁶² used the same methodology for the Gulf Co-operation countries from 1971 to 2002; and also confirmed the growth-led energy consumption hypothesis. Ozturk et al.⁶³ used panel co-integration from 1971 to 2005 for 51 countries classified as low and middle-income countries, and found a long-run Granger-causality running from GDP to energy consumption for low-income countries.

Mehrara⁶⁴ used panel co-integration and causality tests for 11 oil-exporting countries, namely: Iran, Kuwait, UAE, Saudi Arabia, Bahrain, Oman, Algeria, Mexico, Nigeria, Ecuador and Venezuela from 1971 to 2002; and confirmed the growth-led energy consumption hypothesis.

Odhiambo⁶⁵ used the ARDL bounds-testing procedure for Ghana, Cote d’Ivoire, Brazil and Uruguay from 1972 to 2006, and found the growth-led energy consumption hypothesis to hold in the case of Ghana and Cote d’Ivoire only. In 2010, Odhiambo⁹ also used the same approach for the Democratic Republic of Congo, Kenya and South Africa for the period from 1972 to 2006, and found the relationship to hold for the Democratic Republic of Congo. Kayikci and Bildirici⁶⁶ also validated the growth-led energy consumption hypothesis in a study on economic growth and electricity consumption in the Arab states of the Gulf, the Middle East and North African countries from 1972 to 2011. They used the ARDL bounds-testing approach, and found that the direction of the Granger-causality tests differs for each country, depending on the level of their natural resources endowment. Causality was found to flow from GDP to electricity consumption for those countries that do not have enough natural resources, implying that a policy of conserving electricity consumption could be implemented – with little or no effect on economic growth.

Odhiambo⁶⁷ examined the causal relationship between coal consumption and economic growth in South Africa covering the period 1980–2012. Using the ARDL bounds testing approach, he found that there is a unidirectional causal flow from economic growth to coal consumption in South Africa, both in the long run and in the short run. Thus, these results supported the conservation hypothesis. Shahbaz et al.⁶⁸ examined the causality between energy consumption and economic growth in India in the period from 1960 to 2015 and found results in support of the conservation hypothesis, where there is unidirectional Granger-causality from economic growth to energy consumption.

A number of studies have also found the relationship between energy consumption and economic growth to exhibit a bidirectional causal relationship. Odhiambo,⁶⁹ for example, used a trivariate causality framework for South Africa, and found distinct bidirectional causality between electricity consumption and economic growth. Ozturk et al.⁶³ in a panel co-integration framework found bidirectional causality between energy consumption and economic growth for middle-income countries. Al-Mulali and Sab²⁴ found a bidirectional causal relationship between GDP and energy consumption in sub-Saharan Africa in the long run, based on error-correction modelling.

Francis et al.⁷⁰ used BVAR models and co-integration techniques to investigate the energy consumption and economic growth nexus in Haiti, Jamaica and Trinidad and Tobago for the period 1971 to 2002. The study found that in the short run, the feedback hypothesis holds true for all three countries; whereas in the long run, it only holds true for Trinidad and Tobago. Tang⁷¹ used error-correction modelling and ARDL tests for Malaysia, and found the feedback hypothesis to hold true between electricity consumption and economic growth.

Solarin and Shabaz⁷² used ARDL bounds-testing and VECM causality test and confirmed the feedback hypothesis for Angola, between 1971 and 2009. Shahbaz et al.,¹¹ in the case of Pakistan, examined the relationship between renewable energy consumption and economic growth. Using the ARDL model and rolling-window approach, the study confirmed that there is a feedback effect between economic growth and renewable energy consumption.

Naser⁷³ examined the causal linkages between oil consumption and economic growth in four emerging economies – Russia, China, South Korea and India – from 1965 to 2010. The study showed that bidirectional causality exists between oil consumption and economic growth in Russia, China and South Korea. Wang et al.²¹ also investigated the relationship between economic growth, energy consumption, and carbon dioxide emissions in China, from 1990 to 2012. The study employed Granger-causality tests and found a bidirectional causal relationship between economic growth and energy consumption.

Mutascu⁷⁴ investigated the direction of causality between energy consumption and economic growth in G7 countries during the period from 1970 to 2012 using the bootstrap panel causality test. The results were consistent with the feedback hypothesis, where energy consumption and economic growth were mutually causal. Lu⁷⁵ found the same results after utilising the panel cointegration and Granger-causality methodology to examine the energy-growth nexus in Taiwanese industries in the period from 1998 to 2014. Other separate studies that found the feedback hypothesis to exist between energy consumption and economic growth include Adams et al.⁷⁶ and dos Santos Gaspar et al.⁷⁷

Saidi et al.³⁴ investigated the causal relationship between energy consumption and economic growth on a global panel of 53 countries during the period from 1990 to 2014. Using the vector error correction model approach, the results revealed bidirectional Granger causality between economic growth and energy consumption for the global panel both in the short and long runs. In the same year, Carmona et al.⁷⁸ reached the same conclusion in an energy-growth nexus study in the United States.

However, some of the research on the energy consumption and economic growth thesis are consistent with the neutrality view. Altinay and Karagol,⁷⁹ for example, used the Hsiao’s version of Granger-causality test for Turkey, from 1950 to 2000 and validated this hypothesis. In 2007, another study for Turkey by Jobert and Karanfil⁸⁰ also validated the neutrality hypothesis, for the period from 1960 to 2003. Yu and Hwang⁸¹ used the Sims and Granger-causality technique for the United States, from 1947 to 1979; and they found no causal relationship between energy consumption and economic growth.

Using the Granger-causality model, together with the ARDL approach, and Toda and Yamamoto test for New Zealand from 1960 to 1999, Fatai et al.⁸² also found the same relationship between economic growth and energy consumption. Wolde-Rufael⁸³ found no causal relationship in a multi-country study, using Toda and Yamamoto’s Granger-causality test for Benin, Congo, Kenya, Senegal, South Africa, Sudan, Togo, Tunisia and Zimbabwe. Huang et al.⁸⁴ employed a panel VAR modelling approach; and they also found no causal relationship in the case of low-income countries.

Hsiao-Ping¹² validated the neutrality hypothesis for 24 countries from a panel of 49 countries, when investigating oil consumption and output from 1970 to 2010, using the panel-causality analysis. Wolde-Rufael⁸⁵ also confirmed evidence of neutrality between electricity consumption and economic growth in 8 out of 15 transition economies, namely: Albania, Macedonia, Moldova, Poland, Romania, Serbia, the Slovak Republic and Slovenia. The study employed the bootstrap-panel Granger-causality method for the period 1975 to 2010.

Rahman and Mamun⁸⁶ assessed the direction of causality between energy consumption and economic growth in Australia during the period from 1960 to 2012. Using ARDL bounds testing approach, the results were consistent with the neutrality hypothesis, where energy consumption and economic growth were found to be independent of each other and not mutually causal.

Methodology

Although it is now well-appreciated that causality results from a bivariate model might suffer from the omitted variable bias,^87,88 a number of existing studies on energy-growth causality are still based on a bivariate framework. To distinctly differentiate itself from the majority of the related studies, this study employs a multivariate Granger-causality model – which is based on an autoregressive distributed lag (ARDL) bounds-testing approach, as developed by Pesaran and Shin (1999);⁸⁹ and as later enhanced by Pesaran et al.⁹⁰ – to examine the dynamic causal linkage between energy consumption and economic growth in Ethiopia.

In this study, the annual growth rate of real GDP is used as a proxy for economic growth (y). This proxy has been used extensively in literature.^91,92 Energy consumption, on the other hand, is proxied by energy use, as measured by the kilograms of oil equivalent per capita (E). Three additional variables – namely: financial development, investment and trade openness – have been incorporated, as intermittent variables – to form a multivariate Granger-causality model. The choice of these variables as intermittent variables is underpinned by both the theoretical and empirical literature.

The study utilises the causality model that originates from Granger’s definition of causality, based on the notion that the future cannot cause the past; but the past can cause the future. Within the context of this study, using the energy consumption (E) and economic growth (y) variables, the Granger’s definition can be stated as follows: If ‘E causes y’, then changes in E should precede changes in y. In other words, for E to Granger-cause y, two conditions must be met. Firstly, E should help predict y, i.e. in a regression of y against past values of E and y as independent variables; E should contribute significantly to the explanatory power of the regression. Secondly, y should not help to predict E. If E helps to predict y; and y helps to predict E, then it is more likely that one or more variables are in fact, causing both E and y.

The dynamic causal linkage between energy consumption and economic growth in Ethiopia is explored using the recently developed ARDL bounds-testing approach. The approach has a number of advantages over conventional estimation techniques, such as the residual-based technique and the full-maximum likelihood (FML) test.^89,92–94 The first advantage is that it does not impose the restrictive assumption that all the variables need to be integrated of the same order. Thus, the ARDL approach can be utilised to test the existence of a relationship between variables that are integrated of order zero [I(0)] or order one [I(1)], or a mixture of the two.

The second advantage is that the ARDL-based co-integration method provides unbiased long-run estimates and valid t-statistics – even when some of the regressors are endogenous.⁹³ Thirdly, the ARDL approach is based on only a single reduced form equation; while conventional co-integration methods estimate the long-run relationship within the context of a system of equations.⁸⁹ The fourth advantage is that the ARDL approach takes a sufficient number of lags to capture the data-generating process in a general-to-specific modelling framework, in order to obtain optimal lag length per variable. The fifth advantage is that the ARDL procedure possesses superior small-sample properties; hence, it is suitable even when the sample size is small. Based on these advantages, the ARDL approach is considered the most suitable method of analysis in this study. This technique has also been increasingly used in empirical research of late.

The co-integration relationship among the variables is tested, using the ARDL-based co-integration test; and a system of co-integration equations, associated with the multivariate Granger-causality models, is expressed as follows.

ECM-based co-integration model

\begin{array}{l} Δ y_{t} = α_{0} + \sum_{i = 1}^{n} α_{1 i} Δ y_{t - i} + \sum_{i = 0}^{n} α_{2 i} Δ E_{t - i} + \sum_{i = 0}^{n} α_{3 i} Δ F_{t - i} + \sum_{i = 0}^{n} α_{4 i} Δ I_{t - i} \\ + \sum_{i = 0}^{n} α_{5 i} Δ T_{t - i} + α_{6} y_{t - 1} + α_{7} E_{t - 1} + α_{8} F_{t - 1} + α_{9} I_{t - 1} + α_{10} T_{t - 1} + μ_{1 t} \end{array}

(1)

\begin{matrix} Δ E_{t} = β_{0} + \sum_{i = 0}^{n} β_{1 i} Δ y_{t - i} + \sum_{i = 1}^{n} β_{2 i} Δ E_{t - i} + \sum_{i = 0}^{n} β_{3 i} Δ F_{t - i} + \sum_{i = 0}^{n} β_{4 i} Δ I_{t - i} \\ + \sum_{i = 0}^{n} β_{5 i} Δ T_{t - i} + β_{6} y_{t - 1} + β_{7} E_{t - 1} + β_{8} F_{t - 1} + β_{9} I_{t - 1} + β_{10} T_{t - 1} + μ_{2 t} \end{matrix}

(2)

\begin{matrix} Δ F_{t} = δ_{0} + \sum_{i = 0}^{n} δ_{1 i} Δ y_{t - i} + \sum_{i = 0}^{n} δ_{2 i} Δ E_{t - i} + \sum_{i = 1}^{n} δ_{3 i} Δ F_{t - i} + \sum_{i = 0}^{n} δ_{4 i} Δ I_{t - i} \\ + \sum_{i = 0}^{n} δ_{5 i} Δ T_{t - i} + δ_{6} y_{t - 1} + δ_{7} E_{t - 1} + δ_{8} F_{t - 1} + δ_{9} I_{t - 1} + δ_{10} T_{t - 1} + μ_{3 t} \end{matrix}

(3)

\begin{matrix} Δ I_{t} = ϑ_{0} + \sum_{i = 0}^{n} ϑ_{1 i} Δ y_{t - i} + \sum_{i = 0}^{n} ϑ_{2 i} Δ E_{t - i} + \sum_{i = 0}^{n} ϑ_{3 i} Δ F_{t - i} + \sum_{i = 1}^{n} ϑ_{4 i} Δ I_{t - i} \\ + \sum_{i = 0}^{n} ϑ_{5 i} Δ T_{t - i} + ϑ_{6} y_{t - 1} + ϑ_{7} E_{t - 1} + ϑ_{8} F_{t - 1} + ϑ_{9} I_{t - 1} + ϑ_{10} T_{t - 1} + μ_{4 t} \end{matrix}

(4)

\begin{matrix} Δ T_{t} = θ_{0} + \sum_{i = 0}^{n} θ_{1 i} Δ y_{t - i} + \sum_{i = 0}^{n} θ_{2 i} Δ E_{t - i} + \sum_{i = 0}^{n} θ_{3 i} Δ F_{t - i} + \sum_{i = 0}^{n} θ_{4 i} Δ I_{t - i} \\ + \sum_{i = 1}^{n} θ_{5 i} Δ T_{t - i} + θ_{6} y_{t - 1} + θ_{7} E_{t - 1} + θ_{8} F_{t - 1} + θ_{9} I_{t - 1} + θ_{10} T_{t - 1} + μ_{5 t} \end{matrix}

(5)

where y is the growth rate of real gross domestic product (a proxy for economic growth); E is the energy use, as measured by kilograms of oil equivalent per capita (a proxy for energy consumption); F is the share of M2 in GDP (a proxy for financial development); I is the share of gross-fixed capital formation in GDP (a proxy for investment); T is the sum of exports and imports, as a percentage of GDP (a proxy for trade openness). $a_{0}, β_{0}, δ_{0}$ _, $ϑ$ ₀ and $θ_{0}$ are the respective constants; α₁ – α₅, β₁ – β₅, δ₁ – δ_5, $ϑ$ ₁ – $ϑ$ ₅ and $θ$ ₁ – $θ$ ₅ are the respective short-run coefficients; α₆ – α₁₀, β₆ – β₁₀, δ₆ – δ_10, $ϑ$ ₆ – $ϑ$ ₁₀ and $θ$ ₆ – $θ$ ₁₀ are the respective long-run coefficients; Δ is the difference operator; n is the lag length; t is the time period; and μ_it is the white-noise error terms.

ECM-based Granger-causality model

Following the work of Narayan and Smyth,⁹⁵ the ECM-based multivariate Granger-causality model adopted in this study can be expressed as follows⁶⁵

Δ y_{t} = α_{0} + \sum_{i = 1}^{n} α_{1 i} Δ y_{t - i} + \sum_{i = 1}^{n} α_{2 i} Δ E_{t - i} + \sum_{i = 1}^{n} α_{3 i} Δ F + \sum_{i = 1}^{n} α_{4 i} Δ I_{t - i} + \sum_{i = 1}^{n} α_{5 i} Δ T_{t - i} + α_{6} E C M_{t - 1} + μ_{1 t}

(6)

Δ E_{t} = β_{0} + \sum_{i = 1}^{n} β_{1 i} Δ y_{t - i} + \sum_{i = 1}^{n} β_{2 i} Δ E_{t - i} + \sum_{i = 1}^{n} β_{3 i} Δ F_{t - i} + \sum_{i = 1}^{n} β_{4 i} Δ I_{t - i} + \sum_{i = 1}^{n} β_{5 i} Δ T_{t - i} + β_{6} E C M_{t - 1} + μ_{2 t}

(7)

Δ F_{t} = δ_{0} + \sum_{i = 1}^{n} δ_{1 i} Δ y_{t - i} + \sum_{i = 1}^{n} δ_{2 i} Δ E_{t - i} + \sum_{i = 1}^{n} δ_{3 i} Δ F_{t - i} + \sum_{i = 1}^{n} δ_{4 i} Δ I_{t - i} + \sum_{i = 1}^{n} δ_{5 i} Δ T_{t - i} + δ_{6} E C M_{t - 1} + μ_{3 t}

(8)

Δ I_{t} = ϑ_{0} + \sum_{i = 1}^{n} ϑ_{1 i} Δ y_{t - i} + \sum_{i = 1}^{n} ϑ_{2 i} Δ E_{t - i} + \sum_{i = 1}^{n} ϑ_{3 i} Δ F_{t - i} + \sum_{i = 1}^{n} ϑ_{4 i} Δ I_{t - i} + \sum_{i = 1}^{n} ϑ_{5 i} Δ T_{t - i} + ϑ_{6} E C M_{t - 1} + μ_{4 t}

(9)

Δ T_{t} = ϑ_{0} + \sum_{i = 1}^{n} ϑ_{1 i} Δ y_{t - i} + \sum_{i = 1}^{n} ϑ_{2 i} Δ E_{t - i} + \sum_{i = 1}^{n} ϑ_{3 i} Δ F_{t - i} + \sum_{i = 1}^{n} ϑ_{4 i} Δ I_{t - i} + \sum_{i = 1}^{n} ϑ_{5 i} Δ T_{t - i} + ϑ_{6} E C M_{t - 1} + μ_{5 t}

(10)

a_{0}, β_{0}, δ_{0}

ϑ

₀ and

θ_{0}

are the respective constants; α₁ – α₆, β₁ – β₆, δ₁ – δ_6,

ϑ

₁ –

ϑ

₆ and

θ

₁ –

θ

₆ are the respective coefficients; Δ is the difference operator; n is the lag length; t is the time period; and μ_it is the mutually uncorrelated white-noise residuals.

Data sources

This study is based on the annual time-series data, from 1971 to 2013. The data sources for this study are the United Nations Conference on Trade and Development (UNCTAD), the National Bank of Ethiopia and the World Bank Group. The growth rate of real GDP, the share of gross fixed capital formation in GDP, and the ratio of imports and exports to the GDP data were obtained from UNCTAD;⁹⁶ while the nominal GDP and M2 data were sourced from the National Bank of Ethiopia.⁹⁷ The energy-consumption data were collected from the World Bank Group’s databank.³³

Results and discussion

Stationarity tests

Although the ARDL technique does not require pre-testing of the variables for the unit root, the test provides guidance on whether the technique is applicable or not; as it is only appropriate for the variables that are integrated of order one [I(1)] and/or zero [I(0)]. Hence, before co-integration test is conducted, the variables should be first tested for stationarity. For this purpose, the Dickey–Fuller generalised least squares (DF-GLS), Phillips-Perron (PP) and Perron⁹⁸ (PPURoot) unit-root tests are used. The PPURoot is employed to cater for possible structural breaks within the dataset. The results of the stationarity tests for all the variables are presented in Table 1.

Table 1.

Stationarity tests of all variables.

	Stationarity of all variables in levels		Stationarity of all variables in first difference
Variable	Without trend	With trend	Without trend	With trend
Panel 1: Dickey–Fuller generalised least squares (DF-GLS)
y	−4.12***	−5.26***	–	–
E	9.25***	6.37***	–	–
F	−1.74	−0.63	−5.53***	−5.79***
I	−0.16	−2.55	−8.05***	−8.36***
T	−1.26	−1.66	−5.22***	−5.14***

Panel 2: Phillips–Perron (PP)
y	−4.11***	−5.18***	–	–
E	8.39***	6.37***	–	–
F	−1.75	−0.95	−5.59***	−5.80***
I	0.13	−2.55	−8.03***	−8.36***
T	−1.31	−1.80	−5.12***	−5.03***
Panel 3: Perron⁹⁸ (PPURoot)
y	−6.22***	−6.36***	–	–
E	1.04	−2.37	−5.95***	−6.65***
F	−4.61	−5.14	−6.97***	−7.34***
I	−4.44	−4.41	−8.95***	−8.87***
T	−2.74	−2.75	−7.25***	−7.13***

***stationarity at 1% significance level.

See Appendix 1 for the PPURoot break dates.

The results of the stationarity tests reported in Table 1, Panels 1 to 3, confirm that all the variables are either integrated of order zero, or one. This confirmation implies that the ARDL approach to co-integration can be applied.

Bounds F-test for co-integration

The ARDL-based co-integration test is carried out in two steps. The first step involves the determination of the order of lags on the first differenced variables in equations (1) to (5). The second step is the application of the bounds F-test to equations (1) to (5), in order to establish whether a long-run relationship between the variables under study exists or not. In each of the five equations, the null hypothesis of no co-integration is tested against the alternative hypothesis of co-integration.

The calculated F-statistic is matched with the critical values provided by Pesaran et al.⁹⁰ In the event that the calculated F-statistic is above the upper bound level, the null hypothesis of no co-integration is rejected. Hence, it is concluded that the variables in question are co-integrated. On the other hand, if the calculated F-statistic is below the lower-bound level, the null hypothesis of no co-integration is accepted; and it therefore concluded that the variables in question are not co-integrated. Conversely, if the calculated F-statistic falls within the upper- and the lower-bound levels, the results are inconclusive. The results of the bounds F-test for co-integration are presented in Table 2.

Table 2.

Bounds F-test for cointegration.

Dependent variable	Function	F-statistic	Cointegration status
y	F(y\|E, F, I, T)	5.75***	Cointegrated
E	F(E\|y, F, I, T)	8.74***	Cointegrated
F	F(F\|y, E, I, T)	1.99	Not cointegrated
I	F(I\|y, E, F, T)	0.91	Not cointegrated
T	F(I\|y, E, F, I)	1.25	Not cointegrated

Asymptotic critical values
Pesaran et al.,⁹⁰p.300, Table CI(iii)Case III	1%		5%		10%
	I(0)	I(1)	I(0)	I(1)	I(0)	I(1)
	3.74	5.06	2.86	4.01	2.45	3.52

***statistical significance at 1% level.

The co-integration results, presented in Table 2, reveal the existence of two co-integrating vectors. Although the existence of co-integration between the variables suggests that there must be Granger-causality in at least one direction, it does not show the direction of causality between the variables.^13,99 The short-run causality is determined by the F-statistics on the explanatory variables, based on the Wald Test or the Variable Deletion Test. However, the long-run causality is confirmed by the significance and the sign of the coefficient of the error-correction term.

Even though the error-correction term has been included in all the Granger-causality equations (equations (6) to (10)), only those equations in which the null hypothesis of no co-integration is rejected (equations (6) and (7)) will be estimated with an error-correction term.^13,99

There are four expected outcomes regarding the causality between energy consumption and economic growth. The first one is a unidirectional causality flowing from energy consumption to economic growth. The second expected outcome is the unidirectional causal flow from economic growth to energy consumption; while the third one is the bidirectional causality between energy consumption and economic growth. The fourth possible outcome is no causality at all between the two variables of interest.

ECM-based Granger-causality results

Following the establishment of the existence of co-integration in the specified Granger-causality model, the Granger-causality is tested using the ARDL-based technique. The lagged error-correction term is incorporated in the relevant equations (equations (6) and (7)). The results of the causality test are displayed in Table 3.

Table 3.

Results of Granger-causality tests.

Dependentvariable	F-statistics [probability]					ECT_t₋₁[t-statistics]
Dependentvariable	Δy_t	ΔE_t	ΔF_t	ΔI_t	ΔT_t	ECT_t₋₁[t-statistics]
Δy_t	–	2.141[0.102]	2.921*[0.099]	0.004[0.952]	4.11**[0.050]	−1.018***[−6.026]
ΔE_t	5.735**[0.024]	–	9.386***[0.001]	4.696**[0.040]	4.087**[0.050]	−0.978***[6.398]
ΔF_t	4.324**[0.046]	5.100**[0.031]	–	0.626[0.435]	1.725[0.198]	–
ΔI_t	0.621[0.436]	5.624**[0.024]	2.609[0.115]	–	7.175***[0.003]	–
ΔT_t	4.558**[0.040]	0.338[0.565]	1.704[0.200]	7.036***[0.010]	–	–

*statistical significance at 10% level.

**statistical significance at 5% level.

***statistical significance at 1% level.

The empirical results reported in Table 3 show that there is a unidirectional Granger-causality from economic growth to energy consumption in Ethiopia. These results apply, irrespective of whether the estimation is done in the short run or in the long run. The short-run results are supported by the F-statistics of Δy in the energy-consumption function that is statistically significant; while the long-run results are confirmed by the error-correction term (ECM _t _-1) in the same function, which is both negative and statistically significant. These results are consistent with the conservation hypothesis – one of the four hypotheses – postulated in the energy-growth theoretical literature. This hypothesis states that it is the increase in economic development that causes the demand for energy to increase. From the empirical front, the results of this study are supported by the works of Odhiambo,⁶⁵ Jalil and Feridun,⁴⁰ Odhiambo⁶⁷ and Shahbaz et al.⁶⁸ Thus, the results of this study are consistent with both theoretical and empirical literature in the energy-growth field of study, both in the short and in the long run.

These results are robust, statistically and technically acceptable and can be relied upon, given the uniqueness and appropriateness of the methodology adopted. The model, being multivariate, takes away the omission-of-variable bias that comes with bivariate causality models.⁶⁷ In addition, the study has used time series techniques, which take into account country-specific effect, thereby rendering the results of the study more reliable as compared to those found by cross-sectional related methodologies.¹⁰⁰ The results of this study are robust and can be relied on because of the ARDL based methodology utilised, unlike many other studies on the energy-growth causality nexus that utilised the residual-based cointegration related methodologies. The ARDL approach performs better even in smaller samples.¹⁰¹ Furthermore, the ARDL approach corrects for possible endogeneity among the explanatory variables as it provides unbiased long-run estimates and valid t-statistics – even when some of the regressors are endogenous.¹⁰²

The results further show that there is: (i) a bidirectional causality between financial development and economic growth in the short run; (ii) unidirectional Granger-causality flowing from financial development to economic growth in the long run; (iii) bidirectional causality between economic growth and trade openness in the short run; (iv) unidirectional causality from trade openness to economic growth in the long run; (v) bidirectional causality between financial development and energy consumption in the short run; (vi) unidirectional causality from financial development to energy consumption in the long run; (vii) bidirectional causality between investment and energy consumption in the short run; (viii) unidirectional causality from investment to energy consumption in the long run; (ix) bidirectional causality between investment and trade openness in the short run; (x) unidirectional causality from trade openness to energy consumption in the short run; (xi) no causality between investment and economic growth, financial development and investment, and between financial development and trade openness in the short run and in the long run.

Conclusion

This study has explored the causal relationship between energy consumption and economic growth – using the time-series data from Ethiopia during the period from 1971 to 2013. Although there has been a proliferation of empirical literature on the relationship between energy consumption and economic growth in a number of countries, very few country-based studies exist for some sub-Saharan African countries, such as Ethiopia, which is currently transforming its energy sector from overreliance on traditional energy sources to modern energy sources, in order to support its developmental agenda. Moreover, some previous studies suffer from a number of methodological weaknesses, including the omission of variable bias associated with bivariate causality models and the mission of country-specific effects associated with cross-sectional data econometric techniques. The study is fundamentally different from the majority of previous studies on energy-growth causality nexus in that it has used a multivariate framework – with financial development, investment and trade openness as the intermittent variables. The study has also utilised the ARDL bounds testing approach to co-integration and the ECM-based Granger-causality tests to examine this linkage. The results of this study are in support of the conservation hypothesis, as they show that in Ethiopia, there is a distinct unidirectional causal flow from economic growth to energy consumption. These results apply, irrespective of whether the estimation is done in the short run or in the long run. Based on these results, it is possible to predict energy changes in Ethiopia – given the changes in economic growth. However, the predictions are only possible when energy consumption is proxied by kilograms of oil equivalent per capita. The study, therefore, recommends that in Ethiopia, policy-makers should consider expanding their energy-mix options, in order to cope up with the future demand arising from increased economic growth.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Sheilla Nyasha is a Senior Research Fellow in the Department of Economics at the University of South Africa.

Yvonne Gwenhure is a Research Assistant in Macroeconomic Policy Analysis Programme in the Department of Economics at the University of South Africa.

Nicholas M Odhiambo is a Professor and Research Chair of Macroeconomic Policy Analysis Programme in the Department of Economics at the University of South Africa.

Appendix 1

References

Dergiades

Martinopoulos

Tsoulfidis

Energy consumption and economic growth: Parametric and non-parametric causality testing for the case of Greece. Energy Econ 2013; 36: 686–697.

Ahmed

Azam

Causal nexus between energy consumption and economic growth for high, middle and low income countries using frequency domain analysis. Renew Sustain Energy Rev 2016; 60: 653–678.

Erdal

Esengun

The causality between energy consumption and economic growth in Turkey. Energy Policy 2008; 36: 3838–3842.

Lee

Chiu

Nuclear energy consumption, oil prices, and economic growth: Evidence from highly industrialized countries. Energy Econ 2011; 33: 236–248.

Wolde-Rufael

Disaggregated energy consumption and GDP: The experience of Shanghai, 1952–1999. Energy Econ 2004; 26: 69–75.

Apergis

Payne

JE.

Energy consumption and economic growth in Central America: Evidence from a panel cointegration and error correction model. Energy Econ 2009; 31: 211–216.

Iodice

Senatore

Influence of ethanol-gasoline blended fuels on cold start emissions of a four-stroke motorcycle. Methodology and results. SAE Technical Paper 2013-24-0117, 2013.

Iodice

Senatore

Atmospheric pollution from point and diffuse sources in a National Interest Priority Site located in Italy. Energy Environ 2016; 27: 586–596.

Odhiambo

NM.

Energy consumption, prices and economic growth in three SSA countries: A comparative study. Energy Policy 2010; 38: 2463–2469.

10.

Govindaraju

Tang

CF.

The dynamic links between CO₂ emissions, economic growth and coal consumption in China and India. Appl Energy 2013; 104: 310–318.

11.

Shahbaz

Loganathan

Zeshan

et al . Does renewable energy consumption add in economic growth? An application of auto-regressive distributed lag model in Pakistan. Renew Sustain Energy Rev 2015; 44: 576–585.

12.

Hsiao-Ping

Oil consumption and output: What causes what? Bootstrap panel causality for 49 countries. Energy Policy 2012; 51: 907–915.

13.

Odhiambo

NM.

Energy consumption and economic growth nexus in Tanzania: An ARDL bounds testing approach. Energy Policy 2009; 37: 617–622.

14.

Tang

Deng

Zhang

et al . Nexus between energy consumption and economic growth in China: From the perspective of embodied energy imports and exports. Emerg Markets Finance Trade 2016; 52: 1298–1304.

15.

Chiou-Wei

Chen

Zhu

Economic growth and energy consumption revisited – evidence from linear and nonlinear Granger-causality. Energy Econ 2008; 30: 3063–3076.

16.

Kraft

On the relationship between energy and GNP. J Energy Develop 1978; 3: 401–403.

17.

Murray

Nan

GD.

A definition of the gross domestic product- electrification interrelationship. J Energy Develop 1996; 19: 275–283.

18.

Yang

HY.

A note on the causal relationship between energy and GDP in Taiwan. Energy Econ 2000; 22: 309–317.

19.

Ghosh

Electricity consumption and economic growth in India. Energy Policy. 2002; 30: 125–129.

20.

Narayan

Smyth

Electricity consumption, employment and real income in Australia evidence from multivariate Granger-causality tests. Energy Policy 2005; 33: 1109–1116.

21.

Wang

Fang

et al . The relationship between economic growth, energy consumption, and CO₂ emissions: Empirical evidence from China. Sci Total Environ 2016; 542: 360–371.

22.

Tugcu

Ozturk

Aslan

Renewable and non-renewable energy consumption and economic growth relationship revisited: Evidence from G7 countries. Energy Econ 2012; 34: 1942–1950.

23.

Leung

GCK.

Coal consumption and economic growth in China. Energy Policy 2012; 40: 438–443.

24.

Al-Mulali

Sab

CNBC.

The impact of energy consumption and CO₂ emission on the economic growth and financial development in the sub-Saharan African countries. Energy 2012; 39: 180–186.

25.

Eggoh

Bangake

Rault

Energy consumption and economic growth revisited in African countries. Energy Policy 2011; 39: 7408–7421.

26.

Saidi

Hammami

The impact of CO₂ emissions and economic growth on energy consumption in 58 countries. Energy Rep 2015; 1: 62–70.

27.

International Energy Statistics. Key world energy statistics. International Energy Statistics. 2015.

28.

Japan International Cooperation Agency (JICA). Energy policy of Ethiopia. Ministry of Mines & Energy. http://eneken.ieej.or.jp/data/3195.pdf (accessed 7 April 2016).

29.

Ministry of Water and Energy (MoWE). Energy balance and statistics for the years 2005/06 to 2009/10. Addis Ababa: MoWE, 2011.

30.

Gebreegziabher

Hailu

Gebrekirstos

Energy, growth and environmental interaction in the Ethiopian economy. Ethiopia: Mekelle University, 2010.

31.

Guta

Damte

Fe Rede

The residential demand for electricity in Ethiopia. Environment for Development Discussion Paper Series. http://www.rff.org/files/sharepoint/WorkImages/Download/EfD-DP-15-07.pdf (accessed 7 April 2016).

32.

Iodice

Senatore

Industrial and urban sources in Campania, Italy: The air pollution emission inventory. Energy Environ 2015; 26: 1305–1318.

33.

World Bank. World Bank Development Indicators. http://data.worldbank.org (accessed 5 March 2016).

34.

Saidi

Rahman

Amamri

The causal nexus between economic growth and energy consumption: New evidence from global panel of 53 countries. Sustain Cities Soc 2017; 33: 45–56.

35.

Tiba

Omri

Literature survey on the relationships between energy, environment and economic growth. Renew Sustain Energy Rev 2017; 69: 1129–1146.

36.

Shahbaz

Zakaria

Shahzad

SJH

et al. The energy consumption and economic growth nexus in top ten energy-consuming countries: Fresh evidence from using the quantile-on-quantile approach. Energy Economics 2018; 71: 282–301.

37.

Ebohon

OJ.

Energy, economic growth and causality in developing countries: A case study of Tanzania and Nigeria. Energy Policy 1996; 24: 447–453.

38.

Templet

PH.

Energy, diversity and development in economic systems: An empirical analysis. Energy Policy 1999; 30: 223–233.

39.

Ozturk

A literature survey on energy-growth nexus. Energy Policy 2010; 38: 340–349.

40.

Jalil

Feridun

Energy-driven economic growth: Energy consumption-economic growth nexus revisited for China. Emerg Markets Finance Trade 2014; 50: 159–168.

41.

ESH

Choi

JY.

The causal relationship between energy and GNP: An international comparison. J Energy Develop 1985; 10: 249–272.

42.

Bowden

Payne

JE.

The causal relationship between US energy consumption and real output: A disaggregated analysis. J Policy Model 2009; 31: 180–188.

43.

Pao

Tsai

CO2 emissions, energy consumption and economic growth in BRIC countries. Energy Policy 2010; 38: 7850–7860.

44.

Wolde-Rufael

Energy consumption and economic growth: The experience of African countries revisited. Energy Econ 2009; 31: 217–224.

45.

Apergis

Payne

JE.

A panel study of nuclear energy consumption and economic growth. Energy Econ 2010; 32: 545–549.

46.

Shiu

Lam

Electricity consumption and economic growth in China. Energy Policy 2004; 32: 47–54.

47.

Akinlo

AE.

Electricity consumption and economic growth in Nigeria: Evidence from cointegration and co-feature analysis. J Policy Model 2009; 31: 681–693.

48.

Ciarreta

Zarraga

Economic growth and electricity consumption in 12 European Countries: A causality analysis using panel data. Working Paper, Department of Applied Economics III (Econometrics and Statistics), University of the Basque Country. http://www.et.bs.ehu.es/biltoki/EPS/ dt200804.pdfS (accessed 7 April 2016).

49.

Iyke

BN.

Electricity consumption and economic growth in Nigeria: A revisit of the energy-growth debate. Energy Econ 2015; 51: 166–176.

50.

Azam

Khan

Bakhtyar

et al . The causal relationship between energy consumption and economic growth in the ASEAN-5 countries. Renew Sustain Energy Rev 2015; 47: 732–745.

51.

Fang

Chang

Energy, human capital and economic growth in Asia Pacific countries - evidence from a panel cointegration and causality analysis. Energy Econ 2016; 56(C): 177–184.

52.

Destek

MA.

Natural gas consumption and economic growth: Panel evidence from OECD countries. Energy 2016; 114: 1007–1015.

53.

Cheng

Lai

TW.

An investigation of cointegration and causality between energy consumption and economic activity in Taiwan. Energy Econ 1997; 19: 435–444.

54.

Cheng

BS.

Energy consumption, employment and causality in Japan: A multivariate approach. Indian Econ Rev 1998; 33: 19–29.

55.

Zhang

Cheng

Energy consumption, carbon emissions, and economic growth in China. Ecol Econ 2009; 68: 2706–2712.

56.

Stern

Enflo

Causality between energy and output in the long run. Energy Econ 2013; 39: 135–146.

57.

Aqeel

Butt

MS.

The relationship between energy consumption and economic growth in Pakistan. Asia Pacific Develop J 2001; 8: 101–110.

58.

Erol

ESH.

On the causal relationship between energy and income for industrialized countries. J Energy Develop 1987; 13: 113–122.

59.

Lee

CC.

The causality relationship between energy consumption and GDP in G-11 countries revisited. Energy Policy 2006; 34: 1086–1093.

60.

Ang

JB.

Economic development, pollutant emissions and energy consumption in Malaysia. J Policy Model 2008; 30: 271–278.

61.

Ouedraogo

NS.

Energy consumption and economic growth: Evidence from the economic community of West African States (ECOWAS). Energy Econ 2013; 36: 637–647.

62.

Al-Iriani

MA.

Energy-GDP relationship revisited: An example from GCC countries using panel causality. Energy Policy 2006; 34: 3342–3350.

63.

Ozturk

Aslan

Kalyoncu

Energy consumption and economic growth relationship: Evidence from panel data for low and middle income countries. Energy Policy 2010; 38: 4422–4428.

64.

Mehrara

Energy consumption and economic growth: the case of oil exporting countries. Energy Policy. 2007; 35: 2939–2945.

65.

Odhiambo

NM.

Energy dependence in developing countries: Does the level of income matter?

Atl Econ J 2014; 42: 65–77.

66.

Kayicki

Bildirici

Economic growth and energy consumption in GCC and MENA countries. S Afr J Econ 2015; 83: 303–316.

67.

Odhiambo

NM.

Coal consumption and economic growth in South Africa: An empirical investigation. Energy Environ 2016; 27: 215–226.

68.

Shahbaz

Van Hoang

Mahalik

et al . Energy consumption: Financial development and economic growth in India: New evidence from a nonlinear and asymmetric analysis. Energy Econ 2017; 63: 199–212.

69.

Odhiambo

NM.

Electricity consumption and economic growth in South Africa: A trivariate causality test. Energy Econ 2009; 31: 635–640.

70.

Francis

Moseley

Iyare

SO.

Energy consumption and projected growth in selected Caribbean countries. Energy Econ 2007; 29: 1224–1232.

71.

Tang

CF.

examination of the relationship between electricity consumption and economic growth in Malaysia. Energy Policy 2008; 36: 3077–3085.

72.

Solarin

Shahbaz

Trivariate causality between economic growth, urbanisation and electricity consumption in Angola: Cointegration and causality analysis. Energy Policy 2013; 60: 876–884.

73.

Naser

Analysing the long-run relationship among oil market, nuclear energy consumption, and economic growth: Evidence from emerging economies. Energy 2015; 89: 421–434.

74.

Mutascu

A bootstrap panel Granger causality analysis of energy consumption and economic growth in the G7 countries. Renew Sustain Energy Rev 2016; 63: 166–171.

75.

WC.

Electricity consumption and economic growth: Evidence from 17 Taiwanese industries. Sustainability 2016; 9: 50.

76.

Adams

Klobodu

EKM

Opoku

EEO.

Energy consumption: Political regime and economic growth in sub-Saharan Africa. Energy Policy 2016; 96: 36–44.

77.

Dos Santos Gaspar

Marques

Fuinhas

JA.

The traditional energy-growth nexus: A comparison between sustainable development and economic growth approaches. Ecol Indic 2017; 75: 286–296.

78.

Carmona

Congregado

Feria

et al . The energy-growth nexus reconsidered: Persistence and causality. Renew Sustain Energy Rev 2017; 71: 342–347.

79.

Altinay

Karagol

Structural break, unit root, and the causality between energy consumption and GDP in Turkey. Energy Econ 2004; 26: 985–994.

80.

Jobert

Karanfil

Sectoral energy consumption by source and economic growth in Turkey. Energy Policy 2007; 35: 5447–5456.

81.

ESH

Hwang

BK.

The relationship between energy and GNP: Further results. Energy Econ 1984; 6: 186–190.

82.

Fatai

Oxley

Scrimgeour

Energy consumption and employment in New Zealand: Searching for causality. In: NZAE conference, Wellington, New Zealand, 2002.

83.

Wolde-Rufael

Energy demand and economic growth: The African experience. J Policy Model 2005; 27: 891–903.

84.

Huang

Hwang

Yang

CW.

Causal relationship between energy consumption and GDP growth revisited: A dynamic panel data approach. Ecol Econ 2008; 67: 41–54.

85.

Wolde-Rufael

Electricity consumption and economic growth in transition countries: A revisit using bootstrap panel granger-causality analysis. Energy Econ 2014; 44: 325–330.

86.

Rahman

Mamun

SAK.

Energy use, international trade and economic growth nexus in Australia: New evidence from an extended growth model. Renew Sustain Energy Rev 2016; 64: 806–816.

87.

Loizides

Vamvoukas

Government expenditure and economic growth: Evidence from trivariate causality testing. J Appl Econ 2005; 8: 125–152.

88.

Pradhan

Financial development, growth and stock market development: The trilateral analysis in India. J Quant Econ 2011; 9: 134–145.

89.

Pesaran

Shin

An autoregressive distributed lag modelling approach to cointegration analysis. In: Storm S (ed.) Econometrics and economic theory in the 20th Century, The Ragnar Frisch centennial symposium. New York: Cambridge University Press, 1999, pp. 1–31.

90.

Pesaran

Shin

Smith

Bound testing approaches to the analysis of level relationship. J Appl Econ 2001; 16: 174–189.

91.

Shan

Jianhong

Does financial development ‘lead’ economic growth? The case of China. Ann Econ Finance 2006; 1: 231–250.

92.

Majid

MSA.

Does financial development matter for economic growth in Malaysia? An ARDL bound testing approach. J Econ Cooperation 2008; 29: 61–82.

93.

Odhiambo

NM.

Financial depth, savings and economic growth in Kenya: A dynamic causal linkage. Econ Model 2008; 25: 704–713.

94.

Duasa

Determinants of Malaysian trade balance: An ARDL bounds testing approach. J Econ Cooperation 2007; 28: 21–40.

95.

Narayan

Smyth

Energy consumption and real GDP in G7 countries: New evidence from panel cointegration with structural breaks. Energy Econ 2008; 30: 2331–2341.

96.

United Nations Conference on Trade and Development. Economic trends. http://unctadstat.unctad.org (accessed 16 February 2016).

97.

National Bank of Ethiopia. Quarterly bulletin (various issues). http://www.nbe.gov.et/publications/quarterlybulletin.html (accessed 16 February 2016).

98.

Perron

Further evidence on breaking trend functions in macroeconomic variables. J Econom 1997; 80: 355–385.

99.

Narayan

Smyth

Temporal causality and the dynamics of exports, human capital and real income in China. Int J Appl Econ 2004; 1: 24–45.

100.

Ghirmay

Financial development and economic growth in Sub-Saharan African countries: Evidence from time-series analysis. Afr Develop Rev 2004; 16: 415–432.

101.

Tang

TC.

A reassessment of aggregate import demand function in the Asean-5: A cointegration analysis. Int Trade J 2010; 18: 239–268.

102.

Wolde-Rufael

Coal consumption and economic growth revisited. Appl Energy 2010; 87: 160–167.