Does energy-environmental Kuznets curve hold for Ethiopia? The relationship between energy intensity and economic growth

Energy plays critical role in bringing the sustainable development. However, when not used efficiently, its negative consequences on any economy is enormous. This could be related to threatening the sustainable development of the economy, energy security and worsening the environmental conditions (Adom and Kwakwa 2014). Challenges like climate change and energy security that world economy is facing the in the twenty first century is mainly due to energy consumption. Since anthropogenic emissions of carbon dioxide (CO₂) result predominantly from the combustion of fossil fuels, energy consumption is at the focus of the climate change debate. According to U.S Energy Information and Administration (EIA) (2016), world energy-related CO₂ emissions increase from 32.3 billion metric tons in 2012 to 35.6 billion metric tons in 2020 and to 43.2 billion metric tons in 2040.

Efficient energy utilization has a great role in an economy. Promoting energy efficiency has several potential benefits which consist of promoting market competitiveness, saving energy cost, creating a low carbon economy and improving clean energy access. The energy intensity ratio is closely related to the energy efficiency concept (Adom and Kwakwa 2014).

The energy intensity, which is measured as the ratio of energy consumption to the gross domestic product, is one of the most common indicators in the international energy policy evaluations. It has frequently been considered a proxy of a nation’s energy efficiency (Ma and Yu 2017; Metcalf 2008; O’Neill and Chen 2002; Salim et al. 2017; Zhang et al. 2016). Energy intensity is a standard way to measure the link between energy use and economic development (Kander et al. 2017). Moreover, it is a key determinant of the projections of future energy demands (Soni et al. 2017).

Levels of energy intensity remain high in Africa. This situation has raised worries about the possible effect of the current energy consumption patterns on the energy balance in the region’s atmosphere, environmental sustainability, reduction in greenhouse gas emissions and food security (Adom 2015). A continuous decline in energy intensity will determine the economic activities of a country toward a more sustainable path (Ma and Yu 2017). An appropriate selection of policy intervention on energy conservation needs to know the underlying causes of energy intensity because it is both an essential step to energy conservation and a rough basis for projecting the requirements of energy consumption besides the related environmental effects. But, accurate information on the impact of factors such as economic growth on the energy intensity in Africa in general and in Ethiopia in particular has not been studied thoroughly so far.

Ethiopia has large energy resources that need to be exploited to speed up its socio-economic development. Regardless of the presence of a variety of energy resources, the bulk of the national energy consumption is met from biomass energy sources. According to the Ethiopian Environmental Protection Agency (EEPA 2004), about 95% of the total population in Ethiopia uses biomass fuels for their main source of energy. Even though urban households have better access to modern energy than the rural population, the difference in biomass use is not large—approximately 99% of rural households compared to 94% of urban households. The heavy dependence and inefficient utilization of biomass resources have contributed to the depletion of forest resources in Ethiopia (Beyene and Koch 2013).

Due to fast growing economy, energy consumption and production in Ethiopia have been increasing but with a widening gap, leading to an increase in imports and energy insecurity. Along with energy consumption, CO₂ impacts have increased as the energy consumption is mainly from biomass energy source (Ramakrishna 2014). Due to the dependence on biomass for cooking, CO₂ emissions in Ethiopia have increased from 5.1 million tons in 2005 to 6.5 million tons in 2010 (Mondal et al. 2018).

To avoid such negative effects, the government has developed a strategy to build a green economy. The Climate-Resilient Green Economy (CRGE) initiative developed by the country follows a sectoral approach and has so far identified and prioritized more than 60 initiatives, which could help the country achieve its development goals while limiting 2030 GHG emissions to around today’s 150 Mt CO₂e—around 250 Mt CO₂e less than estimated under a conventional development path (Federal Democratic Republic of Ethiopia (FDRE) 2011). Therefore, in achieving this goal, efficient utilization of energy and knowing factors behind it play a crucial role.

The impact of economic growth on energy intensity has been the subject of extensive debate in the energy literature (Csereklyei and Stern 2015). Survey of the empirical literature reveals that empirical studies on the relationship between economic growth and energy intensity can be grouped into three main strands. The first strand of the literature deals with determinants of energy intensity in which economic growth plays a central role (see Filipovic et al. 2015; Jiang and Ji 2016; Jimenez and Mercado 2014; Metcalf 2008; Sadorsky 2013; Song and Zheng 2012). This line of research argues that economic growth has linear positive/negative effect on energy intensity. Evidences show that rising per capita income contributes to declines in energy intensity, primarily through improvements in energy efficiency (see Galli 1998; Jiang and Ji 2016; Mahmood and Ahmad 2018; Metcalf 2008; Sadorsky 2013; Song and Zheng 2012; Yan 2015). Contrary, it is asserted that a higher income increases the demand for energy and hence escalates energy intensity (e.g., Burke and Csereklyei 2016; Rafiq et al. 2016; Sineviciene et al. 2017).

The second theme of the literature focus on testing the validity of Environmental Kuznets Curve (EKC), known as energy-EKC/decoupling or dematerialization hypothesis, in the framework of energy intensity. The energy-EKC hypothesis states that there is inverted U-shape between energy intensity and economic growth. This hypothesis is supported by several empirical studies (e.g., Aboagye 2017; Bilgili 2017; Galli 1998; Jakob et al. 2012; Jiang and Lin 2012). However, the energy-EKC hypothesis has not been established on the solid ground and the empirical findings related to the hypothesis are sensitive to situations like type of variables included and level of economic development. For example, Kander et al. (2017) contended that the clear inverted U-shaped curve in German energy intensity without trade adjustment is reduced when trade is accounted for.

On the other hand, some authors (for example, Lescaroux 2011) criticized that it is wrong to accept a bell shape of energy intensity–economic growth relationship for granted. The shape may be dependent on initial conditions, variables included in the model and the availability of modern commercial energy resources. The energy-EKC hypothesis is sensitive to the inclusion/exclusion of energy embodied in trade (Kander et al. 2017) and energy prices (Richmond and Kaufmann 2006). Moreover, Jakob et al. (2012) claim that energy-EKC hypothesis holds only in high-income countries. There are empirical studies that do not support the energy-EKC hypothesis (Csereklyei and Stern 2015; Jakob et al. 2012; Luzzati and Orsini 2009; Moreau and Vuille 2018). In contrast to the inverted U-shape, some pieces of the literature also suggest an S-shaped relationship between economic growth and energy intensity (Medlock and Soligo 2001; Smil 2003; van Benthem 2015; van Benthem and Romani 2009).

The third strand of the literature deals with testing of convergence in energy intensity across countries. These empirical literatures examine two types of convergences: sigma-convergence and beta-convergence (Deichmann et al. 2018a). Sigma-convergence tests whether energy intensity across counties is getting smaller, while beta-convergence examines whether less efficient countries reduce their energy intensity faster than more efficient ones. Majority of the empirical studies support the convergence hypothesis (see Bulut and Durusu-Ciftci 2018; Burke and Csereklyei 2016; Csereklyei and Stern 2015; Ezcurra 2007; Hajko 2014; Herrerias 2012; Jakob et al. 2012; Jiang et al. 2018; Karimu et al. 2017; Kiran 2013; Liddle 2010; Mulder and de Groot 2012) but, the result is sensitive to different issues. For example, the finding might be affected by structural breaks (Karimu et al. 2017), size of data set (Liddle 2010), types of estimates (Hajko 2014), sector (Jiang et al. 2018), and geographical differences (Liddle 2010); OECD and Eurasian countries show greater convergence than SSA, while convergence is absent in Latin American, Caribbean, North East and North Africa. On the other hand, Pen and Sévi (2010)found that global convergence hypothesis did not hold for a group of 97 countries during 1971–2003 but, local convergence is weakly supported.

This short-lived survey of the literature shows several gaps with the existing literature. Firstly, most global energy models are developed focusing primarily on issues that are important in industrialized countries. Given the fast growing importance of energy trajectories of developing countries for global sustainability, there is a need to develop energy models in the way that can better capture the dynamics of energy systems in developing countries (van Ruijven et al. 2008). Secondly, almost all of the previous studies were conducted using panel data. But, the cross-sectional (or panel) regression model test of β-convergence has been criticized (Bulut and Durusu-Ciftci 2018) because it can only test the hypothesis that whether all countries in the sample are converging or not; therefore, it cannot satisfactorily address country-specific issues (Bernard and Durlauf 1995). Moreover, Evans (1996) argued that cross-country regressions are built on highly unlikely assumptions which can never be satisfied by the real data, and inferences drawn from cross-sectional analyses are misleading because of the inappropriateness of these data for analyzing dynamic behavior of the data (Quah 1993). Therefore, energy consumption dynamics can only be analyzed at the level of individual countries because of heterogeneity of developing countries as argued in Blanchard (1992). Thirdly, the three strands of the literature on the relationship between energy intensity and economic growth presents mixed and inconclusive evidences.

Moreover, to our knowledge, there is no research conducted on testing the existence of energy-EKC hypothesis in Ethiopia so far.

It is against this backdrop that this study examines the impact of economic growth on energy intensity in Ethiopia in the context of energy-EKC hypothesis for the first time, to the level of our knowledge. This article uniquely contributes to the energy literature in that it applies recently developed econometric techniques which have advantageous over the conventional ones. Moreover, robustness of the result has been checked as the EKC hypothesis is too sensitive to the econometric method used.

The rest of the paper is structured as follows. Section 2 of the article outlines methods and data. Section 3 presents findings and discussions of the study. Finally, conclusion and policy implication is presented in Sect. 4.

The estimated result shows that urbanization has positive statistically significant impact on energy intensity. More specifically, a 1% increase in urbanization leads an increment in energy intensity by 2.6%, keeping other things constant. Urbanization may affect energy intensity through various ways. Economic activities are concentrated in cities and metropolitan areas due to urbanization process, and this results in economies of scale in production. This creates opportunity for improving energy efficiency. Urbanization induces the expansion of energy-intensive industries and stimulates the realization of agricultural modernization, which further leads to an increasing energy demand. Urbanization encourages substitution of decentralized traditional wood fuel by centralized modern energy sources. Moreover, urbanization increases the demand for energy because it affects mobility and transport by increasing the amount of motorized traffic into and out of urban areas. Third, increased urbanization increases the demand for infrastructure. Growing cities, for example, increase the demand for energy-intensive products and materials as infrastructure is built. Fourth, urbanization can impact energy demand through its impact on private consumption patterns. Urbanization is accompanied by economic development, and as urban dwellers become wealthier, their consumption patterns change to include more energy-intensive products. Obvious examples include refrigerators, air conditioning and automobiles. The urban centers get easily connected to the national grid and this increases energy consumption. Also, urbanization ignites industrialization. This implies that as the degree of urbanization intensifies, the rate of energy usage increases through the industrialization process. This results in increased use of energy-using appliances like refrigerator. Our finding is in line with Adom and Kwakwa (2014), Rafiq et al. (2016), Sadorsky (2013), Yan (2015)and Jones (1991).

The finding further indicates that import has negative effect on energy intensity in Ethiopia during the specified period. An increase in import by 1% leads to 0.68% decrease in energy intensity, other things remain constant. This is due to the fact that imports of goods and services lead to productivity gains via stronger competition for domestic firms because of the presence of foreign-owned firms and rivaling imports. Imports directly improve productivity, especially if the imported goods have better characteristics than the domestically produced goods. Imports indirectly create productivity spillovers (technology transfer effect) via imitation of the imported products and via improved application of methods adopted together with the imported goods (Aboagye 2016; Adom and Kwakwa 2014; Hübler and Keller 2010). Moreover, import which plays important role in trade openness intensifies economic integration of the economy into the world economies which in turn facilitates the transfer of energy saving technologies. Therefore, import reduces energy intensity through the learning and demonstration (Adom and Kwakwa 2014). Our result is in line with the findings of Belloumi and Alshehry (2016), Adom and Kwakwa (2014) and Adom (2015).

The coefficient of estimated lagged ECT (Table 3) is negative and statistically significant at 1% level of significance which shows that deviation from the long-term energy intensity path due to certain shock is adjusted by 111.95% over the next year. A coefficient of the lagged error-correction term in the short-run model is about − 1.12, which indicates that instead of monotonically converging to the equilibrium path directly, the error-correction process fluctuates around the long-run value in a dampening manner. However, once this process is complete, convergence to the equilibrium path is rapid (Loayza and Ranciere 2005; Narayan and Smyth 2005). All variables under consideration affect energy intensity in Ethiopia in the short-run for the period under study.

3.1 Stability of the model

Misspecification of model in time series data may generate biased results. This situation may diminish the explaining power of the empirical findings (Farhani and Ozturk 2015). Therefore, this study applied the cumulative sum (CUSUM) and cumulative sum squares (CUSUMSQ) of the recursive residuals in order to test for structural stability of the estimated coefficients. Figures 1 and 2 reveal that plots of CUSUM and CUSUMSQ statistics are strictly within the 5% critical bounds indicating that all the estimated coefficients of the error-correction model are stable. Therefore, the selected energy intensity model can be used to understand the decision-making policy regarding energy intensity.

The study examined the impact of economic growth, economic growth squared, import, aid, industrialization and urbanization on energy intensity in Ethiopia using time series data from 1974 to 2014. Clemente, Montanes and Reyes (1998) and Zivot and Andrews (1992) unit root tests which take structural breaks into account in the series were applied in order to produce a result that is not biased toward non-rejection. We applied the ARDL bounds test approach to cointegration and FMOLS methods in order to test the existence of cointegration among the variables under consideration. TY approach to Granger causality was also applied to examine the direction of causal relationship among the variables. To explore the dynamic relationship and timing of energy intensity and its determinants, generalized impulse response and generalized forecast error decomposition were employed.

Analysis based on ARDL and FMOLS approach to cointegration shows the existence of long-run equilibrium relationship among the variables which implies that the variables exhibit common deterministic trend. Economic growth has positive effect on energy intensity. This means that economic growth which a result of increased economic activities leads to higher energy intensity because it requires more energy consumption. On the other hand, economic growth squared has negative and statistically significant effect on energy intensity implying that economic growth at higher stage of development helps reduce energy intensity. The implication is that economic growth has a threshold effect on energy intensity change.

The result also reveals that urbanization has positive impact on energy intensity. This happens due to the fact that urbanization encourages economic activities which are energy intensive. Moreover, urbanization increases energy intensity by increasing amount of motorized traffic into and out of cities, by increasing demand for infrastructure, by changing private consumption patterns and by spurring industrialization.

The result further shows that import has negative and statistically significant impact on energy intensity. Import reduces energy intensity by increasing productivity gains through firms’ competition. It increases productivity, it intensifies economic integration, it creates productivity spillovers (technology transfer effect) and it intensifies economic integration of the economy into the world economies which in turn facilitates the transfer of energy saving technologies.

The finding of the study has strong theoretical and policy implications. Theoretically, the fact that economic growth affects energy intensity positively while economic growth squared has negative effect on energy intensity replicates the argument of Eenergy-Environmental Kuznets Curve (Eenergy-EKC) which corroborates with works of Deichmann et al. (2018b), Zhang et al. (2016) and Filipovic et al. 2015). The positive impact of urbanization on energy intensity was corroborated by the findings of Sadorsky (2013), Belloumi and Alshehry (2016) and Sadorsky (2013). The finding that shows import has negative energy intensity reflects the fact that international openness that are reflected in FDI, and trade openness contributed to the reduction in energy intensity (Adom 2015; Adom and Kwakwa 2014; Jiang and Ji 2016).

As policy implication, Ethiopian energy policy should focus on energy conserving strategies. More specifically, economic growth of Ethiopia which has been taking place at an alarming rate requires higher energy consumption. Currently, Ethiopia’s energy consumption is predominantly dominated by biomass (traditional) energy sources which are not efficient as compared to the modern one. Therefore, policymakers should device policies through which the country’s energy intensity can be reduced. More specifically, the country’s energy source should be shifted from traditional energy sources to the modern ones. Energy conservation strategies should be developed; energy efficiency should be increased; and utilization of cleaner energy should be increased. In addition, the government of Ethiopia should implement sound environmental policies which encourage the import of cleaner and energy saving (efficient) technologies for reducing energy intensity. As far as urbanization is concerned, the government should pay due attention to public infrastructure at phase of design, construction and operation to be more energy saving and efficient. Moreover, transportation technologies should be energy efficient and the government should pay due attention to improve public awareness of energy saving so as to change the urban residents’ consumption pattern and lifestyle.

Future research should focus on the impact of trade structure, foreign direct investment, energy price, structural change and the economic integration on energy intensity. Moreover, cointegration tests that consider several structural breaks should be employed to advance the knowledge relating to factors affecting energy intensity in the case of Ethiopia.