. International Journal of Energy Economics and Policy | Vol 7 • Issue 3 • 2017102 International Journal of Energy Economics and Policy ISSN: 2146-4553 available at http: www.econjournals.com International Journal of Energy Economics and Policy, 2017, 7(3), 102-109. The Relationship between Energy Consumption, Economic Growth and Carbon Dioxide Emission: The Case of South Africa Hlalefang Khobai1*, Pierre Le Roux2 1Department of Economics, Nelson Mandela Metropolitan University, South Africa, 2Department of Economics, Nelson Mandela Metropolitan University, South Africa. *Email: hlalefangk@gmail.com ABSTRACT This paper investigates the relationship between energy consumption, carbon dioxide (CO2) emission, economic growth, trade openness and urbanization for South Africa. The annual data for the period between 1971 and 2013 is employed. The results of Johansen test of co-integration show that there is a long run relationship between energy consumption, CO2 emission, economic growth, trade openness and urbanization in South Africa. The results for the existence and direction of vector error correction model (VECM) Granger causality indicates that there is bidirectional causality flowing between energy consumption and economic growth in the long run. The VECM results further found a unidirectional causality flowing from CO2 emissions, economic growth, trade openness and urbanization to energy consumption and from energy consumption, CO2 emissions, trade openness and urbanization to economic growth. These results posit a fresh perspective for creating energy policies that will boost economic growth in South Africa. Keywords: Energy Consumption, Economic Growth, Carbon Dioxide Emission, South Africa JEL Classifications: O13, Q43 1. INTRODUCTION The world has experienced major changes in economic and environmental scopes in South Africa due to the great reform and economic transition in the past two decades. South Africa has experienced increasing demand in energy following the increase in economic growth post-apartheid era. The country’s energy intensity is above average which indicates that much energy is required to produce a single unit of gross domestic product (GDP). However, South Africa’s energy utilization is characterized by high dependence on low-cost and abundantly available coal. A large amount of crude oil is imported into the country while a small amount of renewable energy is used. Table 1 shows the trends of the total energy supply from 2003 to 2006. It can be seen from Table 1 that coal dominates the energy supply while hydroelectricity contributes the least to energy supply. However, coal was subjected to a mix of trends since 2003. Its contribution decreased by 4.5% from 2003 to 2004, and increased by 3.6% from 2004 to 2005; from 2005 to 2006, it fell again by 5.9%. In general, the contribution of coal has decreased by 6.8% for the entire period. The contribution of hydroelectricity has increase by 0.1% since 2003. Crude oil has experienced an increase of 7.8% since 2003 while gas supply increased by 1.7%. The nuclear contribution dropped from 3.1% in 2003 to 1.9% in 2006. The renewables have decreased from 9.4% in 2003 to 7.6% in 2006. Coal and crude oil remain the major primary energy suppliers in South Africa despite their effects on air quality, human health, wildlife and climate change. South Africa is the ranked number six among the world’s largest recoverable coal reserves (Department of Energy, 2009). It is 12th highest carbon dioxide (CO2) emitter in the world and number one greenhouse gas emitter in Africa (USAID, 2016). The increasing concern of the greenhouse gas emission has motivated many researchers to investigate relationship the between energy consumption, CO2 emission and economic growth in different countries and regions. However, this relationship has been rarely examined in South Africa despite the fact that South Africa’s energy consumption and carbon emission increased more than double in the last two decades. While there are studies carried in the international literature to investigate Khobai and Le Roux: The Relationship between Energy Consumption, Economic Growth and Carbon Dioxide Emission: The Case of South Africa International Journal of Energy Economics and Policy | Vol 7 • Issue 3 • 2017 103 the relationship between energy consumption, CO2 emission and economic growth, they do not focus on South Africa (Saidi and Hammami, 2015; Arouri et al., 2012; Linh and Lin, 2014; Vidyarthi, 2013). The studies which were done in South Africa focused on energy consumption and economic growth (Okafor, 2012; Wolde-Rufael, 2009 and Odhiambo, 2010). The only South African study that covered energy consumption, economic growth and pollutant emissions was done by Menya and Wolde-Rufael (2010). Our study differs from Menya and Wolde-Rufael (2010)’s study in that we included trade openness and urbanization as the additional variables to form a multivariate framework. The remainder of the study is organized as follows: Section 2 studies the review of the empirical literature. Section 3 presents the data and methodology used in the study followed by the discussion of the findings in Section 4. Section 5 concludes the study with some policy implications. 2. LITERATURE REVIEW The relationship between economic growth and energy consumption had been researched extensively over the past three decades. However, the results concerning the direction of causality between these variables are still mixed. The results range from no causality to unidirectional or bidirectional causality. The difference is caused by different methodologies applied, different countries’ data applied and the particular period of the study. The pioneers of the studies on energy consumption and economic growth are Kraft and Kraft (1978). Their study considered the case of USA for the period 1947 to 1974. The results from Sims Granger-causality supported a unidirectional causality flowing from gross national product to energy consumption. This implies that energy conservation policies can be introduced without causing any harm to economic growth. Shyamal and Rabindra (2004) undertook a study to investigate the causal relationship between energy consumption and economic growth in India covering the period 1950-1996. The study utilized Engle-Granger co-integration to estimate the long run relationship between these variables and the standard Granger-causality to find the direction of causality. The results supported a unidirectional causality flowing from energy consumption to economic growth. The results from Engle-Granger co-integration detected a one-way causality flowing from economic growth to energy consumption in the long term. The combination of standard Granger-causality and the Engle-Granger approach revealed bidirectional causality between energy consumption and economic growth. Saidi and Hammami (2015) conducted a study to assess the link between energy consumption and economic growth in Tunisia. The annual data was used for the period between 1974 and 2011. The Johansen technique results suggested that there is a long run relationship between economic growth and energy consumption. The Granger-causality results found a bidirectional causality flowing between energy consumption and economic growth in Tunisia. Tang et al. (2016) investigated the long run relationship between energy consumption and economic growth for the period between 1971 and 2011. The results from co-integration technique revealed existence of co-integration among the variables. The Granger- causality results suggested a one-way causality running from energy consumption to economic growth in Vietnam. This implies that Vietnam is an energy-intensive country and there is a need to implement renewable energy policy to provide sufficient supply as this will speed up economic growth. Albiman et al. (2015) conducted a study to determine the relationship between energy consumption, environmental pollution and per capita economic growth in Tanzania for the period between 1975 and 2013. The study investigated the causality relationship by employing the more robust causality technique of Toda and Yamamoto’s non-causality test. The findings revealed a unidirectional causality flowing from economic growth and energy consumption to environmental pollution through CO2 emissions. Vidyarthi (2013) carried out a study to investigate the long term and causal relationship between energy consumption, economic growth and carbon emissions in India. The data used in this study covered a period from 1971 to 2009. To determine the co-integration between the selected variables, the Johansen co-integration technique was employed while the vector error correction model (VECM) Granger-causality test was used to find the direction of causality between the variables. The Johansen co- integration technique results established a long term relationship between energy consumption, carbon emissions and economic growth. The long term causality results validated a unidirectional causality flowing from energy consumption and CO2 emissions to economic growth while the short term causality revealed mixed results: A unidirectional causality flowing from energy consumption to carbon emission; carbon emission to economic growth and; economic growth to energy consumption. Table 1: Total primary energy supply-TJ: 2003-2006 Variables 2003 % 2004 % 2005 % 2006 % Coal 3,227,600 72.7 3,573,343 68.2 3,651,726 71.8 3,721,156 65.9 Crude Oil 615,689 13.7 1,016,664 19.4 724,774 14.2 1,214,122 21.5 Gas 50,218 1.1 84,152 1.6 153,078 3.0 160,318 2.8 Nuclear 138,142 3.1 145,801 2.8 123,193 2.4 109,375 1.9 Hydro 2890 0.1 2890 0.1 4199 0.1 11,069 0.2 Nuclear 422,979 9.4 418,058 8.0 430,427 8.5 428,396 7.6 Total 4,507,518 5,240,908 5,089,397 5,644,436 Source: Department of Energy (2009) Khobai and Le Roux: The Relationship between Energy Consumption, Economic Growth and Carbon Dioxide Emission: The Case of South Africa International Journal of Energy Economics and Policy | Vol 7 • Issue 3 • 2017104 Another study that used a trivariate framework is by Dehnavi and Haghnejad (2012) which aimed to determine the relationship between energy consumption, pollution and economic growth for a selected panel of eight Organization of the Petroleum Exporting Countries. The study used a panel data technique for the period 1971-2008. The Granger-causality identified a two-way causality between CO2 emissions and energy consumption and a one-way causality flowing from economic growth to energy consumption and pollution in the long term. The short run results observed that economic growth Granger-causes CO2 emission while energy consumption Granger-cause CO2 emission and economic growth. A multi-country study was conducted by Vidyarthi (2014) to investigate the relationship between energy consumption, carbon emissions and economic growth for five South Asian countries: Bangladesh, India, Pakistan, Nepal and Sri Lanka. The study used Pedroni’s co-integration to determine the long term relationship among the variables and panel VECM Granger-causality to find the direction of causality between the variables. In using data for the period between 1972 and 2009, the study found that there exists a long term relationship between energy consumption, carbon emissions and economic growth in all these countries. The VECM Granger-causality suggested bidirectional causality between energy consumption and economic growth, a unidirectional causality flowing from carbon emissions to economic growth and energy consumption in the long term. The short term results identified a unidirectional causality flowing from energy consumption to carbon emissions. Another multi-country study was carried out by Ahmed and Azam (2016) who served to examine the nexus between energy consumption and economic growth for 119 countries all over the world. The countries were categorized as follows; 30 high income OECD, 13 high income non-OECD, 65 middle income and 11 low income countries. The Granger-causality results detected feedback hypothesis for 18 countries (5 high income OECD, 2 high income non-OECD, 10 middle income OECD and 1 low income). It was further established that there is a growth hypothesis in 25 countries which comprises of 4 high income OECD, 3 high income non-OECD, 14 middle income and 4 low income countries. conservation hypothesis was revealed for 6 high income OECD, 6 high income non-OECD, 27 middle income and 1 low income countries. finally, there was no causality established in 15 high income OECD, 2 high non-OECD, 14 middle income and 5 low income countries. Moubarak and Lin’s (2014) research aimed to determine the long term and short run relationship between renewable energy consumption and economic growth in China. The data used in the study covered the period 1977-2011. CO2 emissions and labor were used as additional variables to form a multivariate framework. The results from the autoregressive distributed lag technique and the Johansen co-integration test found that there is a long term relationship between the selected variables. The Granger-causality test suggested a two-way causality flowing between renewable energy consumption and economic growth in the long-term. The causality results further established a unidirectional causality flowing from labor to renewable energy consumption. There was no causality found between carbon emissions and renewable energy consumption in the long term and short term. This implies that renewable energy has not been exploited in China to mitigate CO2 emissions. Pablo-Romero and De Jesús (2016) conducted a study to examine the link between energy consumption and economic growth using the hypothesis postulated for the energy-environmental Kuznets curve. A panel data of 22 Latin American and Caribbean countries were employed covering the period between 1990 and 2011. Their findings showed existence of a U-shaped relationship between energy consumption and economic growth. Linh and Lin (2014) contribute to the most recent studies that assessed the dynamic relationship between energy consumption and economic growth using multivariate framework by adding the variables foreign direct investments and CO2 emissions. This Vietnam study used data for the period between 1980 and 2010. The co-integration findings show that there is a long term relationship between economic growth, energy consumption, foreign direct investments and CO2 emissions. The Granger- causality results established bidirectional causality between foreign direct investment and income in Vietnam. This implies that an increase in Vietnam’s income has a potential of attracting more capital from overseas. Kais and Sami (2016) investigated the impact of energy consumption and economic growth on CO2 emissions in 58 countries covering the period between 1990 and 2012. The countries were divided into three regional subgroups as follows: European and North Asian region, Latin American and Caribbean region and Sub-Saharan region. The results posit that energy consumption has a positive impact on CO2 emissions all panels. It is further revealed that economic growth has a positive and a statistically significant impact on CO2 emissions European and North Asian region and North Africa and Sub-Saharan Africa. Streimikiene and Kasperowicz (2016) served to determine the nexus between energy consumption and real GDP by incorporating fixed capital and total employment to form multivariate framework. The study used data for 18 European Union countries for the period from 1995 to 2012. It was established that economic growth, energy consumption and gross fixed capital move along in the long run. The findings from the fully modified ordinary least squares (FMOLS) and dynamic ordinary least squares estimators indicated a positive relationship between economic growth, energy consumption and gross fixed capital. Arouri et al. (2012) examined the link between energy consumption, CO2 emission and economic growth for 12 Middle East and North African countries spanning the period 1981-2005. The bootstrap panel unit root tests and co-integration techniques were applied. The findings indicated a long run positive relationship between energy consumption and CO2 emissions. Furthermore, real GDP showed a quadratic relationship with CO2 emissions for the entire region. Saidi and Hammami’s (2015) study serves to investigate the impact of economic growth and CO2 emissions on energy consumption for Khobai and Le Roux: The Relationship between Energy Consumption, Economic Growth and Carbon Dioxide Emission: The Case of South Africa International Journal of Energy Economics and Policy | Vol 7 • Issue 3 • 2017 105 58 countries. Dynamic panel data model for the period between 1990 and 2012 was estimated using the generalized method of moments (GMM). The findings show that CO2 emission have a positive and significant impact on energy consumption for four global panels. The results further show that economic growth has a positive effect on energy consumption. Wang et al. (2016) detected the causal link between urbanization, energy consumption and CO2 emissions for the Association of Southeast Asian Nations countries covering the period between 1980 and 2009. The findings from the Pedroni panel co-integration tests evidenced existence of a long run relationship among the variables. Employing the FMOLS, it was established that all else the same, a 1% rise in urbanization leads to CO2 emissions increasing by 0.20%. The Granger-causality results suggested a unidirectional short run causality flowing from urbanization to energy consumption and from urbanization to CO2 emissions. It was further detected that urbanization and energy consumption Granger-cause CO2 emissions in the long run. Ozturk and Al-Mulali (2015) investigate whether better governess and corruption control help to form the inverted U-shaped relationship between income and pollution in Cambodia for the period of 1996-2012. The outcome from the GMM and the two-stage least squares revealed that GDP, urbanization, energy consumption, and trade openness increase CO2 emission while the control of corruption and governess can reduce CO2 emission. It is fundamental to note that the environmental Kuznets curve hypothesis was not confirmed in Cambodia. Okafor (2012) and Odhiambo (2010) employed energy consumption and economic growth nexus in South Africa. Odhiambo (2010) applied the Granger-causality test while Okafor (2012) employed Hsiao’s Granger-causality test. The results of Odhiambo’s (2010) study validated a one-way causality flowing from energy consumption to economic growth while Okafor’s (2012) results suggested a unidirectional causality flowing from energy consumption to economic growth. Menyah and Wolde-Rufael’s (2010) research investigated the relationship between economic growth, pollutant emissions and energy consumption. Their study added labor and capital to form a multivariate model and used South African data for the 1965-2006 period. A modified version of the Granger- causality test and bounds test approach to co-integration were applied to analyze the direction of causality and long term relationships between the variables. A long term relationship was established between the variables. The Granger-causality results showed unidirectional causality flowing from pollutant emissions to economic growth. It also found a unidirectional causality from energy consumption to CO2 emission and from energy consumption to economic growth. Al-Mulali et al. (2015) investigate the influence of disaggregated renewable electricity production by source on CO2 emission in 23 selected European countries for the period of 1990-2013. The Pedroni cointegration results indicated that CO2 emission, GDP growth, urbanization, financial development, and renewable electricity production by source were cointegrated. Moreover, the fully modified ordinary least-square results revealed that GDP growth, urbanization, and financial development increase CO2 emission in the long run, while trade openness reduces it. From the empirical literature, it can be realized that no study was done to investigate the relationship between energy consumption, economic growth and CO2 emission incorporating trade openness and urbanization as controlling variables in South Africa. Therefore, this current study endeavors to fill that gap. 3. METHODOLOGY 3.1. Model Specification This study analyses the relationship between energy consumption, CO2 emissions and economic growth by adding trade openness and urbanization as intermittent variables to form a multivariate framework. To address the issue of heteroskedasticity, all the variables are converted into logarithm form. The log linear quadratic form is used to analyze the relationship between energy consumption, CO2 emissions and economic growth using the following model; LEC LCO LGDP LTO LUBN t LCO t LGDP t LTO t LUBN t t = + + + + + α α α α α ε 1 2 2 (1) LEC represents the natural log of energy consumption per capita LGDP indicates the natural log of real GDP (using constant prices of 2010), LCO2 is the natural log of CO2 emissions, LTO denotes natural log trade openness and LUBN represents natural log of urbanisation. Furthermore, α1 and εt represent the constant and an error term, respectively. 3.2. Data Sources The study employs annual time-series data covering the period between 1971 and 2013 for energy consumption, economic growth, CO2 emissions, trade openness and urbanization. Different sources have been used to gather data of the mentioned variables. Real GDP (using constant prices of 2010) was collected from the South African Reserve Bank. The data for CO2 emissions, energy consumption and urbanization were collected from Word Development Indicators while data for trade openness was sourced from United Nations and Trade Development. 3.3. Data Analysis 3.3.1. Unit root test The unit root test is used to determine whether or not the variables energy consumption, CO2 emission, economic growth, trade openness and urbanization are stationary series. This study employs two unit root tests; augmented Dickey–Fuller (ADF) unit root test by Said and Dickey (1984) and another one by Phillips and Perron (1988) termed Phillips-Perron (PP) unit root test. When the variables are found to be integrated of the same order, co-integration between the variables will be tested. 3.3.2. Co-integration test The long run relationship between energy consumption, CO2 emissions, economic growth, trade openness and urbanization is estimated using the Johansen co-integration technique (Johansen, 1988; Johansen and Juselius, 1990). This technique involves the estimation of a VECM to determine the likelihood-ratios. It works Khobai and Le Roux: The Relationship between Energy Consumption, Economic Growth and Carbon Dioxide Emission: The Case of South Africa International Journal of Energy Economics and Policy | Vol 7 • Issue 3 • 2017106 in a way that there are at most n−1 cointegrating vectors if there are n variables which all have unit roots. The VECM model employed in this study is as follows: ∆ = + ∆ + ∝ + = − −∑ −Y Yt i k i t Y t t kθ θ β ε 0 1 1 1 ' (2) Where, Δ is the difference operator, Yt is (LEC, LCO2, LGDP, LTO, LUBN), θ is stands for the intercept and ε is the vector of white noise process. The Johansen technique comprises of two likelihood ratio tests namely; the maximum eigenvalue and the trace test. The number of co-integrating vectors in the system is determined by the number of significant non zero Eigen values. 3.3.3. Granger-causality The presence of co-integration implies that there is a causality but it does not show the direction of causality among the variables. To estimate causality between energy consumption, CO2 emission, economic growth, trade openness and urbanization, the VECM is employed. The empirical equations of the VECM Granger- causality are presented as follows: ∆ ∆ ∆ ∆LEC LEC LCO LGDPt t i t i t i i s i r i q = + + +− − − === ∑∑α α α α10 11 12 2 13 111 ∑∑ ∑ ∑+ + + +− − − − −α α ψ ε14 1 15 1 1 1 1 ∆ ∆LTR LUBN ECTt i i t t i i u t t (3) ∆ ∆ ∆ ∆ LCO LCO LEC LGDP t t i t i i r i q t i 2 20 21 2 22 11 23 2 = + + + + − − == − ∑∑α α α α α 44 11 25 1 2 1 2 ∆ ∆ LTR LUBN ECT t i i t i s t i i u t t − −= − − − ∑∑ ∑+ + +α ψ ε (4) ∆ ∆ ∆ ∆ LGDP LGDP LEC LCO t t i t i i r i q t i = + + + + − − == − ∑∑α α α α α 30 31 32 11 33 2 344 11 35 1 3 1 3 ∆ ∆ LTR LUBN ECT t i i t i s t i i u t t − −= − − − ∑∑ ∑+ + +α ψ ε (5) ∆ ∆ ∆ ∆LTR LTR LEC LCOt t i t i t i i s i r i q = + + +− − − === ∑∑∑α α α α40 41 42 43 2 111 ++ + + +− − − − −∑ ∑α α ψ ε44 1 45 1 5 1 5 ∆ ∆LGDP LUBN ECTt i i t t i i u t t (6) ∆ ∆ ∆ ∆ LUBN LUBN LEC LCO t t i t i i r i q t i i = + + + + − − == − = ∑∑α α α α 50 51 52 11 53 2 11 54 1 55 1 5 1 5 s t i i t t i i u t t LGDP LTR ECT ∑ ∑ ∑ − − − − −+ + + α α ψ ε ∆ ∆ (7) Where, ECt, CO2t, GDPt, TRt, UBNt, represent energy consumption, CO 2 emissions, GDP, trade openness and urbanization, respectively. εit (for i = 1, 2, 3, 4, 5) represents serially uncorrelated random error terms. ECTt−1 (error correction term) represents the co-integrating vectors. The adjustment coefficient is ψ and it shows how much disequilibrium is corrected (Jamil and Ahmed, 2010). To find the long term causality flowing from the dependent variable(s) to the dependent variable, the coefficient of the ECT (ψ) should be significant. From equation 3, the causality from CO2, GDP, TR, UBN to EC can be tested. From equation 4, the causality from EC, GDP, TR, UBN to CO2 can be estimated while from equation 5, the causality from EC, CO2, TR, UBN to GDP can be tested. The causality from EC, CO2, GDP, UBN to TR can be estimated from equation 6 and from equation 57, the causality from EC, CO2, GDP, TR to UBN can be tested. The Wald test on differenced and lagged differenced terms of the dependent variables is employed to estimate the short run causality. 4. FINDINGS 4.1. Unit Root Tests The time series properties of the variables are tested using the ADF test by Dickey and Fuller (1984) and PP test by Phillips–Perron (1988). The results are presented in Table 2. The results at level form show that all the five variables are non- stationary at the 5% level of significance, whereas at first difference the variables are stationary. This implies that energy consumption, economic growth, CO2, trade openness and urbanization are integrated of order one. 4.2. Co-integration The presence of a similar order of integration, as reported by ADF and PP unit root tests, endorses the application of the Johansen co-integration test. But prior to estimating co-integration among the variables, the optimum lag is determined. The results are Table 2: Unit root tests Variables ADF unit root test PP unit root test Levels 1st difference Levels 1st difference LCO2 −1.6661 −5.9022* −1.8363 −5.8988* LENC −1.9875 −6.2656* −1.8922 −6.2663* LGDP −1.7242 −4.4481* −1.0698 −4.2660* LTO −2.1026 −5.4221* −2.0581 −5.4427* LUBN −1.7675 −2.0591*** −1.3083 −1.3742*** Source: Own calculation. *,***Represent 1% and 10% significance levels, respectively. ADF: Augmented Dickey–Fuller, PP: Phillips–Perron, CO2: Carbon dioxide, GDP: Gross domestic product Khobai and Le Roux: The Relationship between Energy Consumption, Economic Growth and Carbon Dioxide Emission: The Case of South Africa International Journal of Energy Economics and Policy | Vol 7 • Issue 3 • 2017 107 illustrated in Table 3. The study selects the optimum lag to be 2, according to the Schwarz information criterion and Akaike information criterion. Co-integration among the variables is explored using the Johansen co-integration test and the results are presented in Table 4. The Johansen co-integration test exhibits that for r = 0, the λ max statistics is 36.135, which is greater than the 95% critical value of 33.877. On the other hand, maximal Trace statistics is 79.054, which is greater than the 95% critical value of 69.819. This implies that the null hypothesis r = 0 is rejected at 5% level of significance. But the results for R ≤ 1, R ≤ 2, R ≤ 3 and R ≤ 4 shows that the null hypotheses cannot be rejected. As a result, the trace test and the maximum Eigen test detected the existence of a single co- integrating vector. Therefore, the study concludes that there is a long run relationship between energy consumption, CO2 emissions, economic growth, trade openness and urbanization in South Africa. 4.3. Granger-causality The direction of causality between the variables is estimated using the VECM and the findings of both long and short run causalities are presented in Table 5. When energy consumption was used as the dependent variable, the lagged error term was found to negative and significant. This shows that there is one-way causality flowing from CO2 emissions, economic growth, trade openness and urbanization to energy consumption in the long run. Similar results were established when economic growth was used as the dependent variable. This implies that there is existence of a one way causality flowing from energy consumption, CO2 emissions, trade openness and urbanization. Generally, it can be realized that there is bidirectional causality flowing between energy consumption and economic growth. This shows increasing economic growth is essential for the improvement of the energy industry which in turn helps boost economic growth in South Africa. The short run results exhibit no short run causality flowing between energy consumption, economic growth, CO2 emissions, trade openness and urbanization. The absence of a short run causality flowing from energy consumption to economic growth means that environmentally friendly policies such as energy conservation, efficiency improvements measures and demand-side management policies can be implemented in South Africa without adversely affecting economic growth. 4.4. Variance Decomposition Tables 6-8 present variance decomposition results for CO2 emissions, energy consumption and economic growth, respectively. Table 6 illustrates that in the 10th year, one standard deviation shock in energy consumption, economic growth, trade openness and urbanization, reveals 8.10%, 6.97%, 11.45% and 4.61% of the forecast error variance of CO2 emissions, respectively. A greater percentage of 68.88 of variation in economic growth becomes self-explanatory after 10 periods. The variance decomposition approach findings in Table 7 posit that a 62.09% portion of energy consumption is contributed by its own innovative shocks. A one standard deviation shock in CO2 emission explains energy consumption by 13.06% while economic growth, trade openness and urbanization support energy consumption by 9.08%, 12.09% and 3.68%, respectively. Table 3: Selection order criteria Lag LogL LR FPE AIC SC HQ 0 402.7047 NA 1.59e-15 −19.88523 −19.6741 −19.80890 1 724.3525 546.8014 5.82e-22 −34.71763 −33.4509 −34.25964* 2 756.1016 46.03614* 4.42e-22* −35.05508* −32.7329* −34.21544 3 771.8888 18.94460 8.24e-22 −34.59444 −31.2167 −33.37315 Source: Own calculation. *indicates lag order selected by the criterion. LR: Sequential modified LR test statistic (each test at 5%), FPE: Final prediction error, AIC: Akaike information criterion, SC: Schwarz information criterion, HQ: Hannan-Quinn information criterion Table 4: Johansen co-integration H1: Alternative hypothesis H0: Null hypothesis λmax test λmax test (0.95) Trace test Trace test (0.95) R=1 R=0 36.135 33.877 79.054 69.819 R=2 R≤1 16.262 27.584 42.919 47.856 R=3 R≤2 12.232 21.132 26.658 29.979 R=4 R≤3 9.661 14.256 14.425 15.495 R=5 R≤4 4.764 3.841 4.764 3.841 Source: Own calculation Table 5: VECM Dependent variable Types of causality Short run Long run ∑ΔLGDP ∑ΔLENC ∑ΔLCO2 ∑ΔLTO ∑ΔLUBN ECTt−1 ΔLGDP ……. 0.3189 0.1391 0.1027 1.0345 −0.005* ΔLENC 0.0086 ………. 0.1464 0.1783 0.3882 −0.066* ΔLCO2 0.6474 0.1554 …………. 0.2462 1.8034 0.170* ΔLTO 2.5025 1.5289 0.6362 ………. 0.8510 0.003 ΔLUBN 0.3142 0.6291 0.3651 0.1978 ………. −0.106 Source: Own calculation. CO2: Carbon dioxide, GDP: Gross domestic product, ECT: Error correction term, VECM: Vector error correction model. *Represent 1% significance level, respectively Khobai and Le Roux: The Relationship between Energy Consumption, Economic Growth and Carbon Dioxide Emission: The Case of South Africa International Journal of Energy Economics and Policy | Vol 7 • Issue 3 • 2017108 Table 8 illustrate the results of variance decomposition of economic growth. The results exhibit that only 3.83% of the variation in economic growth is self-explanatory, while CO2 emissions accounted for a larger forecast error variance. CO2 emission explain 78.60% of the forecast error variance after 10 periods. The other variables account for the remaining percentages: Energy consumption (7.27%), trade openness (3.00%), and urbanization (7.30%). 5. CONCLUSION The study analyses the relationship between energy consumption, economic growth and CO2 emissions by incorporating trade openness and urbanization as the control variables to form a multivariate framework. The Johansen co-integration technique and the VECM were used to estimate the long run relationship and the direction of causality among the variables. The findings of the Johansen co-integration test demonstrate an existence of one co-integrating equation. This shows that there is a long run relationship between energy consumption, economic growth, CO2 emission, trade openness and urbanization in South Africa. The VECM results detect bidirectional causality flowing between energy consumption and economic growth. This shows that an energy-led growth hypothesis exists in South Africa. The results further proved existence of a one-way causality flowing from CO2 emissions, trade openness and urbanization to economic growth and energy consumption. The findings of the variance decomposition analysis show that the share of energy consumption in explaining economic growth is minimal. The results of this study indicate that policies aiming at reducing energy consumption and controlling for CO2 emissions in South Africa could slow down growth. This implies that any energy conservation measures undertaken should consider the adverse impact on economic growth. South Africa has been found to be one of the highest CO2 emitters in the world. It is therefore important that in finding ways of proving energy services, South Africa Table 6: Variance decomposition of CO2 emissions Period SE CO2 EC GDP TO UBN 1 0.019628 100.0000 0.000000 0.000000 0.000000 0.000000 2 0.026562 91.89339 4.589901 2.555510 0.156643 0.804559 3 0.030665 85.57634 7.246649 3.083585 1.711180 2.382247 4 0.033072 81.11732 7.874258 2.950823 4.377137 3.680463 5 0.034440 77.54523 8.121085 2.866657 7.083195 4.383833 6 0.035192 74.98575 8.241193 2.966109 9.158480 4.648468 7 0.035647 73.16568 8.270247 3.379795 10.47951 4.704775 8 0.036017 71.67708 8.243090 4.221844 11.17391 4.684075 9 0.036394 70.26940 8.180441 5.471839 11.43646 4.641854 10 0.036795 68.87592 8.099492 6.966451 11.44830 4.609839 CO2: Carbon dioxide, GDP: Gross domestic product Table 7: Variance decomposition of energy consumption Period SE EC CO2 GDP TO UBN 1 0.015500 77.83988 22.16012 0.000000 0.000000 0.000000 2 0.020594 82.72739 14.97301 1.759080 0.098657 0.441873 3 0.023183 78.77644 14.16628 2.991576 2.271403 1.794303 4 0.024914 74.54362 14.05594 3.103013 5.430711 2.866717 5 0.025942 71.44868 13.85322 3.219925 8.080770 3.397404 6 0.026544 69.00323 13.72646 3.641936 10.01734 3.611028 7 0.026959 67.00582 13.60893 4.468539 11.24390 3.672818 8 0.027317 65.26040 13.45389 5.732366 11.87785 3.675495 9 0.027673 63.62673 13.26384 7.340313 12.10238 3.666734 10 0.028034 62.09049 13.05784 9.086054 12.08886 3.676760 CO2: Carbon dioxide, GDP: Gross domestic product Table 8: Variance decomposition of economic growth Period SE GPD EC CO2 TO UBN 1 0.009458 3.194818 12.13791 84.66727 0.000000 0.000000 2 0.014698 1.655402 9.026458 87.64497 0.526282 1.146883 3 0.018176 1.123524 8.384503 85.92335 1.553330 3.015294 4 0.020480 0.884926 8.298630 83.73153 2.353602 4.731316 5 0.022015 0.850080 8.192801 82.10033 2.838178 6.018612 6 0.023039 1.127241 8.004994 80.94399 3.083956 6.839818 7 0.023719 1.715172 7.777171 80.10167 3.148813 7.257179 8 0.024172 2.477627 7.560666 79.46727 3.103363 7.391073 9 0.024471 3.232140 7.389231 78.97630 3.031177 7.371148 10 0.024663 3.830934 7.274707 78.59727 2.998905 7.298182 CO2: Carbon dioxide, GDP: Gross domestic product Khobai and Le Roux: The Relationship between Energy Consumption, Economic Growth and Carbon Dioxide Emission: The Case of South Africa International Journal of Energy Economics and Policy | Vol 7 • Issue 3 • 2017 109 should pay attention to the environmental impact associated with different uses of energy. It is recommended that policy makers should advocate for renewable energy such as wind and solar to ensure sustainable growth of the economy. REFERENCES Ahmed, M., Azam, M. (2016), Causal nexus between energy consumption and economic growth for high, middle and low income countries using frequency domain analysis. Renewable and Sustainable Energy Reviews, 60, 653-678. Albiman, M.M., Suleiman, N.N., Baka, H.O. (2015), The relationship between energy consumption, CO2 emissions and economic growth in Tanzania. International Journal of Energy Sector Management, 9(3), 361-375. Al-Mulali, U., Ozturk, I., Lean, H.H. (2015), The influence of economic growth, urbanization, trade openness, financial development, and renewable energy on pollution in Europe. Natural Hazards, 79(1), 621-644. Arouri, M.E., Youssef, A., Mhenni, H., Rault, C. (2012), Energy consumption, economic growth and CO2 emissions in Middle East and North African countries. Energy Policy, 45, 342-349. Dehnavi, J., Haghnejad, A. (2012), Energy consumption, economic growth, and pollution in selected OPEC countries: Testing the environmental Kuznets curve hypothesis. Journal of Academic Research Economics, 4(2), 149-167. Department of Energy. (2009), Energy information management. Digest of South African Energy Statistics. Pretoria, South Africa: Process Design and Publications. Jamil, F., Ahmed, P. (2010), The relationship between electricity consumption, electricity prices and GDP in Pakistan. Energy Policy, 38, 6016-6025. Johansen, S. (1988), Statistical analysis of cointegration vectors. Journal of Economics Dynamic and Control, 12, 231-254. Johansen, S., Juselius, K. (1990), Maximum likelihood estimation and inference on cointegration with applications to the demand for money. Oxford Bulletin of Economics and Statistics, 52, 169-210. Kais, S., Sami, H. (2016), An econometric study of the impact of economic growth and energy use on carbon emissions: Panel data evidence from fifty eight countries. Renewable and Sustainable Energy Reviews, 59, 1101-1110. Kraft, J., Kraft, A. (1978), On the relationship between energy and GNP. Journal of Energy Development, 3, 401-403. Linh, D.H., Lin, S. (2014), CO2 emissions, energy consumption, economic growth and FDI in Vietnam. Managing Global Transitions, 12(3), 219-232. Menyah, K., Wolde-Rufael, Y. (2010), Energy consumption, pollutant emissions and economic growth in South Africa. The Journal of Energy Economics, 32, 1374-1382. Moubarak, M., Lin, B. (2014), Renewable energy consumption - Economic growth nexus in China. Renewable and Sustainable Energy Reviews, 40, 111-117. Odhiambo, N.M. (2010), Energy consumption, prices and economic growth in three SSA countries: A comparative study. Journal of Energy Policy, 38, 2463-2469. Okafor, H.O. (2012), Testing the relationship between energy consumption and economic growth: Evidence for Nigeria and South Africa. Journal of Economics and Sustainable Development, 3(11), 111-124. Ozturk, I., Al-Mulali, U. (2015), Investigating the validity of the environmental Kuznets curve hypothesis in Cambodia. Ecological Indicators, 57, 324-330. Pablo-Romero, M.D.P., De Jesús, J. (2016), Economic growth and energy consumption: The energy-environmental Kuznets curve for Latin America and the Caribbean. Renewable and Sustainable Energy Reviews, 60, 1343-1350. Phillips, P.C.B., Perron, P. (1988), Testing for unit roots in time series regression. Biometrika, 75, 335-346. Said, S.E., Dickey, D. (1984), Testing for unit roots in autoregressive moving-average models with unknown order. Biometrika, 71, 599-607. Saidi, K., Hammami, S. (2014), Energy consumption and economic growth nexus: Empirical evidence of Tunisia. American Journal of Energy Research, 2(4), 81-89. Shyamal, P., Rabindra, N.B. (2004), Causality between energy consumption and economic growth in India: A note on conflicting results. Journal of Energy Economics, 26, 977-983. Streimikiene, D., Kasperowicz, R. (2016), Review of economic growth and energy consumption: A panel cointegration analysis for EU countries. Renewable and Sustainable Energy Reviews, 59, 1545-1549. Tang, C.F., Tan, B.W., Ozturk, I. (2016), Energy consumption and economic growth in Vietnam. Renewable and Sustainable Energy Reviews, 54, 1506-1514. USAID. (2016), Greenhouse Gas Emissions in South Africa. Available from: http://www.users/cqwanzcho/downloads/ghg%20 emissions%20fact%20sheet%20south%20africa%20-%207-5- 16%20usaidsacomm_rev08-26-16_clean%20(1). [Last accessed on 2017 Feb 20]. Vidyarthi, H. (2013), Energy consumption, carbon emissions and economic growth in India. World Journal of Science, Technology and Sustainable Development, 10(4), 278-287. Vidyarthi, H. (2014), An econometric study of energy consumption, carbon emissions and economic growth in South Asia: 1972-2009. World Journal of Science, Technology and Sustainable Development, 11(3), 182-195. Wang, Y., Chen, L., Kubota, J. (2016), The relationship between urbanization, energy use and carbon emissions: Evidence from a panel of Association of Southeast Asian Nations (ASEAN) countries. Journal of Cleaner Production, 112, 1368-1374. Wolde-Rufael, Y. (2009), Energy consumption and economic growth: The experience of African countries revisited. Journal of Energy Economics, 31, 217-224.