TX_1~AT/TX_2~AT International Journal of Energy Economics and Policy | Vol 12 • Issue 1 • 2022 427 International Journal of Energy Economics and Policy ISSN: 2146-4553 available at http: www.econjournals.com International Journal of Energy Economics and Policy, 2022, 12(1), 427-435. The Effects of Gross Domestic Product and Energy Consumption on Carbon Dioxide Emission in Uganda (1986-2018) Jacob Otim1, Geoffrey Mutumba1*, Susan Watundu2, Geoffrey Mubiinzi3, Milly Kaddu1,3 1Department of Economics and Statistics, Kyambogo University, Uganda, 2Department of Management Science, Makerere University Business School, Uganda, 3Department of Economics, Uganda Martyrs University, Nkozi, Uganda. *Email: gmutumba@kyu.ac.ug Received: 29 September 2021 Accepted: 17 December 2021 DOI: https://doi.org/10.32479/ijeep.12552 ABSTRACT This study examines the effects of energy consumption and per capita gross domestic product on carbon dioxide emission which is a precursor for global warming due to its large scale impact on the environment. The effect of per capita gross domestic product and per capita energy consumption on carbon emission per capita in Uganda is not clearly known. This study fills the empirical gap for Uganda for 1986-2018. The study used Vector Error Correction techniques and the results suggest evidence of a long-run relationship between the variables at a 5% significance level using the Johansen cointegration test. The estimated elasticity of carbon dioxide emission per capita with respect to gross domestic product per capita is 1.856. The results for the existence and direction of Granger causality show a unidirectional causality running from gross domestic product per capita to carbon dioxide emission per capita and the environmental Kuznets curve hypothesis is supported. In addition, there is no causal link between energy consumption per capita and gross domestic product per capita, which supports the growth neutrality hypothesis. The overall results indicate that gross domestic product per capita has a positive effect on carbon dioxide emission in Uganda while energy consumption does not Granger cause carbon dioxide emission. Keywords: CO2 emission, energy consumption, GDP per capita, Johansen cointegration test, Granger causality, VECM JEL Classifications: K32, P18, Q43, Q48, Q54 1. INTRODUCTION Global warming has attracted considerable debate in the area of environment in the last three decades (Held and Rogers 2018). This has been attributed to the risk of change in climate on human beings and the ecosystem. Due to large-scale impact of human activities, most countries of the United Nations have agreed to keep the global warming below 2°C by cutting down Greenhouse Gases (GHGs) emission mainly due to CO2. The commitment ratified through the signatures of the Kyoto Protocol in 1997 (Protocol, 1997). This was followed by the Paris Agreement with the main aim of addressing the global response to the threat of climate change (Horowitz, 2016). Carbon emission is the release of carbon to the atmosphere. The emission is composed of GHGs which are the main contributors to climate change. GHGs are often calculated as carbon equivalent. The Kyoto protocol spelled out six main GHG pollutants which have significant impact on the environment namely; CO2, CH4 (Methane), N2O (Nitrous Oxide), HFCs (hydrofluorocarbons), PFCs (perfluorocarbons) and SF6 (sulfur hexafluoride). CO2 is considered to be the main contributor to global warming (Zhang and Da, 2015). The summit on sustainable development held in 2002 in Johannesburg, South Africa highlighted the detrimental impact of energy consumption on the environment despite its key role in economic growth and development (Sghari and Hammami, 2016). In Uganda, CO2 emission from hydrocarbon combustion and industrial use contribute roughly to 0.099% of the global carbon stock. Even though Uganda contributes less to the potentially catastrophic accumulation of man-made carbon footprints, the This Journal is licensed under a Creative Commons Attribution 4.0 International License Otim, et al.: The Effects of Gross Domestic Product and Energy Consumption on Carbon Dioxide Emission in Uganda (1986-2018) International Journal of Energy Economics and Policy | Vol 12 • Issue 1 • 2022428 country is susceptible to the influence of climate change. Uganda’s carbon stock is among the lowest in the world and is estimated at 1.39 tons of carbon equivalent less than the world average of 7.99 tons of carbon emission equivalent per capita (GOU, 2015). However, the contribution of carbon emissions to the growth trajectory of Uganda is not clearly known (Markandya et al., 2015). The Uganda’s vision 2040 may be severely hampered by climate factors in the absence of adaptation actions (GOU, 2020) due to overreliance on traditional biomass as the main source of energy (Bamwesigye et al., 2020). Therefore, the achievement of Sustainable Development Goal (SDG) 7 (affordable clean energy) geared towards ensuring access to affordable, reliable and sustainable modern energy for all and SDG 13 (climate action) may be hampered. Actions to combat climate change and its impacts on the environment would all be practically impossible unless serious policy measures are put in place (Walters, 2021). In this paper, we examine the effects of energy consumption per capita and per capita GDP on CO2 emission per capita, in Uganda using the Johansen cointegration testing approach and vector error correction model (VECM) based on Granger causality for Uganda over 1986-2018 period. The major contribution of this paper is to provide the theoretical understanding of the effect of GDP and energy consumption on the environment in a multivariate framework using a vector error correction model. The rest of the paper is organized as follows. The next section presents the literature review. The third section shows the methods. The fourth section reports the empirical results and discussion. The final section is the conclusion of the study. 2. LITERATURE REVIEW The link between carbon emission, energy consumption and economic growth has been widely analyzed and is a center of controversy and debate since the 1950s (Kuznets, 1955). Climate and energy are intrinsically linked. The quality of our environment, by and large, is determined the way we consume energy. As a result, examining the productive use of energy is key to the sustainable development goals 11 (develop sustainable cities and communities) and 13 (climate change). The awareness of the change in climate and its implication on planet earth makes it vital to examine the causal effect of energy consumption on development. Energy consumption is key in the development process since it is a main driver for sustainable development (Ma, 2019). The rise in energy consumption is expected to lead to the growth in GDP in real terms through a transmission mechanism that cascades to CO2 emission increase, which is an important factor in global warming and eventual climate change. Studying nexus among these global variables is pivotal. There are numerous stock pollutants that lead to climate change but CO2 is a dominant gas of all GHGs (Sadik-Zada and Loewenstein, 2020). Investigating a causal link between carbon emission, energy consumption and economic growth has become a landmark of recent studies since energy consumption is the best vehicle to achieve sustainable development (Akadiri, 2019). Several studies have used a multivariate framework to examine the causal link between CO2 emission, energy consumption, and economic growth, they find mixed results. Alam (2011) uses a Johansen test and autoregressive distributed lag to explain a long-run relationship between electricity consumption, energy consumption, CO2 emission, and GDP in Bangladesh. His results show a one-way causality running from energy consumption to economic development both in the short-run and the long-run. Additional results show a bidirectional causality between CO2 emission and energy consumption and also electricity consumption and economic development. In Chang‘s (2010) study, different energy sources- crude oil, natural gas, electricity and coal are used to examine the relationship between CO2 emission, energy consumption, and GDP in China. The results show that more CO2 emission is a caused by energy consumption and GDP growth. Ang (2007) employs a multivariate error correction model to investigate the causal link between energy consumption, CO2 emission and output in France. The results show energy consumption to increase emissions, while the study by Halicioglu (2009) shows that there is bidirectional causality between output and CO2 emission both in the short-run and long-run for Turkey. A couple of studies have been conducted in sub-Saharan Africa (SSA) Menyah and Wolde-Rufael’s (2010) study, show causality between economic development, energy consumption, and pollutant emission in South Africa and there exist long-run relationships between the variables. The result also indicates that there is one-way causality running from both energy consumption and carbon emission to economic development. Adom et al. (2012) find a bidirectional relationship between CO2 emission and economic growth in Morocco and one-way causality running from economic growth to CO2 emission in Senegal. A study by Appiah (2018) on the multivariate Granger causality between energy consumption, economic growth and CO2 emission in Ghana from 1960-2015 using Toda-Yamamoto and Granger causality test shows a feedback Granger causality between energy consumption and CO2 emission. Further still, a unidirectional Granger causality from energy consumption to economic growth was detected. Studies in Uganda’s context, include Senkantsi and Okot (2016) on electricity consumption- and GDP for Uganda from 1982-2013 using ARDL bound test and the Granger causality framework. Their study found a unidirectional causal flow from economic growth to electricity consumption. Mawejje and Mawejje (2016) also conducted a similar study using vector error correction techniques and Granger causality test, where they found a unidirectional causal link running from electricity consumption to GDP. Appiah et al. (2019) studied the causal link between industrialization, energy intensity, and GDP and carbon oxide emission using data from Uganda from 1990 to 2014. The existence of unidirectional causality comes from energy intensity to CO2 and from GDP to CO2. This implies that the increase in energy intensity leads to more environmental pollution. A similar study that investigates the effect of energy on CO2 emission in Uganda is by Appiah et al. (2019). They use energy intensity as a predictor for CO2 emission in Uganda yet its applications in real world policy making is troublesome. Due to the fact that the GDP of the individual economies are converted to a common currency Otim, et al.: The Effects of Gross Domestic Product and Energy Consumption on Carbon Dioxide Emission in Uganda (1986-2018) International Journal of Energy Economics and Policy | Vol 12 • Issue 1 • 2022 429 using the purchasing power parity or the market exchange rates (Samuelsson, 2014), energy intensity becomes a poor predictor of CO2 emission. The weakness in Appiah et al. (2019) paper gives a justification for our paper to make a theoretical contribution and use energy consumption per capita as a better predictor for CO2 emission in Uganda. 3. METHODS 3.1. Research Design A correlational research design and quantitative approach for time series analysis were adopted. All the data from the selected variables are continuous in nature over time and statistical methods of measurement are used (Beard et al., 2020). The study covered a period of 33 years from 1986 to 2018. This period has been selected because there has been a stable government under one leadership without any serious interruption from war or takeover of government as always common in developing countries. 3.2. Expected Signs of the Variables The expected signs, symbols, measurements, and data sources used in the model are presented in Table 1. The data used for this paper obtained from the World Bank Development indicators. 3.3. Model Specifications Following the empirical literature in economics, we have observed that energy is a significant factor of CO2 emission. According to the EKC hypothesis CO2 emission and output have a nonlinear quadratic relationship (Dinda 2004). At a steady state, the cointegrating relationship between carbon emissions, energy and output can be specified as follows: c ec gdpt t t t� � � �� � � � (1) Where Ct=ln(CO2t/Lt) is per capita CO2 emission, ect=ln(ECt/Lt) is per capita energy consumption, gdpt=ln(GDPt/Lt) is per capita GDP and εt is the residuals. Lt is the total population over time. Equation one is expressed in the natural logarithmic form to reduce the effect of heteroskedasticity and to obtain the growth rate of the relevant variables by differencing their natural logarithmic form. The expected sign in equation 1 is such that θ>0 because high energy consumption is expected to increase CO2 emission. The parameters θ, and γ are long-run elasticities of CO2 emission per capita with respect per capita energy use, and per capita GDP respectively. 3.4. Unit Root Tests The stationarity was tested using the Augmented Dickey-Fuller (ADF) and Philips-Perron (PP) developed by Phillips and Perron (1988). Dickey and Fuller (1981) examine the unit root in each time series with the following hypothesis: Ho: θ=0, when the time series has unit root H1: θ<0, when the time series has no unit root Following the ordinary least square (OLS) assumption, ADF is expressed as: � �y y y ut t t i i p t i t� � � � �� � ��� � � �1 1 (2) Where t is a deterministic trend, ψ and β are the constants, p is the lag order selection based Schwartz Bayesian Criterion (SBC). If the calculated value in the absolute term, is more than the t-statistic, we reject the H0 (θ=0), if H0 is rejected it implies the series is stationary and is an I(0) process. When H0 is rejected at the first difference, the series is I(1). Similarly, when the series becomes stationary at the second difference, the series exhibits an I(2) process. The PP test is used to correct for any serial correlation and heteroskedasticity in the error ut since it is more robust than Dickey-Fuller. 3.5. Cointegration Test We applied the Johansen cointegration test by Johansen and Juselius (1990). It is used to check for existence long-run cointegrating equation(s) between or among variables of the I(1) series. The cointegration is performed in levels but not in first difference. But since the variables are in natural logarithmic form, the log transformed variables are used to interrogate the long-run relationship. The multivariate cointegration model of Johansen and Juselius is expressed as: � � �y y y ut t i t i P t� � � �� � � � �� � 1 1 1 1 (3) Where, π and Гi are coefficient matrices, Δ is the difference operator and P is the lag order selected based on SBC. Johansen and Juselius cointegration uses two likelihood ratio test-the trace and max eigenvalue tests and they are computed as follows: 1 ˆ( ) ln(1 ) n i i r T r T λ = + = − −∑ (4) ( ) ( )max 1ˆ, 1 ln 1 rr r Tλ λ ++ = − − (5) Where, îλ is the expected eigenvalue of the characteristic roots and T is the sample size. In λtrac test, the H1 investigates the number Table 1: Variable description and expected signs Variables Symbol Measure Expected Sign Data source Per capital CO2 emission Ct CO2 emission in Kiloton N/A World bank: World development indicators Per capita GDP GDPt GDP constant 2010 US$ + World bank: World development indicators Per capita Energy consumption ECt Energy consumption measured in Terra Joules + World bank: Sustainable energy for all. Otim, et al.: The Effects of Gross Domestic Product and Energy Consumption on Carbon Dioxide Emission in Uganda (1986-2018) International Journal of Energy Economics and Policy | Vol 12 • Issue 1 • 2022430 of cointegrating vector r against H1 of n cointegrating vector. Whereas, for λmax test, H0 investigates the number of cointegrating vectors r against H1 of r+1 cointegrating vector. When using Johansen and Juselius cointegration test, where there is one or more cointegrating vectors, there is evidence of long-run equilibrium between or among variables. To establish the relationship we set the null hypothesis as follows: H0: No cointegrating equation(s) The decision criteria is based on 5% level of significance. We reject the null if the value of λtrac and λmax is greater than 5% critical value. Otherwise, we fail to reject the null. 3.5. Vector Error Correction Model (VECM) VECM is a system of vector of two or more variables that are exogenous. If the variables are non-stationary but I(1) time series and they are cointegrated, we can run VECM to examine both short-run and long-run dynamics of the series. The VECM for all the endogenous variables appears as follows: � � � � c c ec gdp EC t i i k t i j t j j k m t m m k � � � � � � � � � � � � � � � � � � � � � � 1 1 1 1 1 1 1 TT ut t� �1 1. . (6) � � � � ec c ec gdp t i i k t i j t j j k m t m m k � � � � � � � � � � � � � � � � � � � � � � 1 1 1 1 1 1 2 . EECT ut t� �1 21. (7) � � � � gdp c ec gdp t i i k t i j t j j k m t m m k � � � � � � � � � � � � � � � � � � � � � � 1 1 1 1 1 1 3 EECT ut t� �1 3 (8) Where k−1 the lag length is reduced by 1 and αi, βj, μm are the short-run coefficients of the model’s adjustment long-run equilibrium. λi is the coefficient of ECT and is the speed of the adjustment parameter towards long-run equilibrium. It has to be negative and statistically significant for it to maintain its economic interpretation. ECTt−1 = the term error correction relates to the fact that last year period deviation from long-run equilibrium (error) influences the short-run dynamics of the dependent variables. uit= residuals (stochastic error term often called impulses or innovations or shocks. To find the long-term causality flowing from the dependent variable(s) to the dependent variable, the coefficient of the ECT ( λ ) should be significant and is defined as: ECT c c ec gdpt i i k t j j k t j m m k t m� � � � � � � � � � � � � �� � �� � � �� � � � 1 1 1 1 1 1 1 (9) 3.6. Granger Causality Test We used a two-steps procedure from Engle and Granger (1987) model to investigate the link between CO2 emissions per capita, energy consumption per capita, and GDP per capita. In the first step, we estimate the long-run model in equation 6 in order to get the estimated residuals. The second step is to estimate error correction based on the Granger causality approach. The error correction based on causality tests allows the inclusion of the lagged error term derived from the cointegration equation (Ozturk and Acaravci 2010). We also validated our result using the pairwise Granger causality test. 4. RESULTS AND DISCUSSIONS 4.1. Unit Root Tests Most variables may not exhibit -stationarity, therefore we test for unit root. The time series properties of the variables under investigation are tested using Augmented Dickey-Fuller (ADF) test developed by Dickey (1981) and Philips-Perron (PP) tests by Phillips and Perron (1988) at a constant intercept without a trend are applied. This is always done to avoid spurious results in the data generation process. Table 2 illustrates the results. From Table 2 all variables are non-stationary in levels and they are all stationary at first difference i.e, I(1). Since all the variables are I(1) series, we proceed to test for the existence of the cointegration equation. To do this we determine the appropriate lag length. 4.2. Optimal Lag Length Structure The key issue in time series analysis is the optimal lag selection process for a finite set of observations (Hamilton, 2020). The selection of a lag structure in VECM is an empirical question because both over-fitting and under-fitting of the model with lags result into insignificant coefficients and spurious outcomes. For the purpose of this study, we consider Final prediction error (FPE), Akaike information criterion (AIC), Schwarz information criterion (SC) and Hannan-Quinn information criterion (HQ) for the selection of the optimal lag structure. Table 3 shows the results; the optimal lag length selected by the criteria is one. This is appropriate because we are dealing with annual observations for 33 years. 4.3. Cointegration Test The existence of long-run cointegration is tested using Johansen cointegration test developed by Johansen (1988). It was carried out to establish whether the series are linearly related. In such a situation if there are shocks in the short-run, which often affect the individual series, they can convergein the long-run. If the series are not integrated we only estimate Vector Auto Regressive (VAR) model. To use the Johansen cointegration test, our null hypothesis supports the existence of no cointegrating equation and the decision criteria is based at a 5% level of significance. We reject the null hypothesis if the value of the trace and max statistics is greater than 5% critical value. Otherwise we fail to reject the null hypothesis. Table 4 depicts the Johansen unrestricted integrated rank test. We reject the null hypothesis of no cointegrating equation since the value of Trace statistic (λtrace) and Max-Eigen (λmax) value are above the critical value at 5%. We conclude that there are Otim, et al.: The Effects of Gross Domestic Product and Energy Consumption on Carbon Dioxide Emission in Uganda (1986-2018) International Journal of Energy Economics and Policy | Vol 12 • Issue 1 • 2022 431 long-run cointegration equation. This implies that the series are related and can be combined in a linear fashion and in case there is a shock in the short-run, which may affect the movement in the individual series they would converge in the long-run. Therefore, we can proceed a head to estimate the long-run equation in the Vector Error Correction (VEC) Model framework, because there is at least one cointegrating equation confirmed by the analysis of both Trace and Max-Eigen statistics. The normalise Johansen cointegration results from our estimation is expressed in Table 5. The figures in the parentheses are standard errors. Thus the Johansen long-run cointegration equation is expressed as follows: c ec gdpt t t� � � �18 613 0 526 1 856. . . (10) The result in equation 10 indicates that economic growth measured in terms of real GDP per capita is positively and statistically is associated with CO2 emission. While energy consumption is positively corrected with CO2 emission but it is not statistically significant. Specifically, a 1% increase in GDP per capita tends to increase CO2 emission by about 1.856%, ceteris paribus. We conclude that the null hypothesis of no cointegration is rejected against its alternative of cointegrating relationship in the model. The positive link between CO2 emission and GDP per capita makes rational economic sense, since the growth in GDP per capita is associated with the increase in the economic activities through consumption of goods and services which are likely to have negative effect of the environment. However, there is a lot of debate on this area in terms of causality between economic growth and CO2 emission with no consensuses reach yet among the scholars as shown in the literature section. 4.4. Granger Causality and Short-run Dynamics This paper also explores the direction of causality between the variables by using error-correction based Granger causality models which are weak (short-run) Granger causality and long-run causality. The results for models in Tables 6-8 can be summarised as follows: (i) There is no causality between CO2 emission per capita and energy consumption. (ii) A unidirectional relationship running from GDP per capita to CO2 emissions per capita and supports EKC hypothesis in the short-run. (iii) There is no causality running from energy consumption per capita to GDP per capita. (iv) There is no causality from GDP per capita to energy consumption per capita. (v) Long-run relationship exists for CO2 emission per capita equation and GDP per capita Granger cause CO2 emission per capita. The findings of positive relationship between economic growth and carbon emissions are in line with Appiah et al. (2019) in Uganda’s context. Sekantsi and Okot (2016) and Mawejje and Mawejje, (2016) but did not test the causality between CO2 emissions and economic growth. The generally, the results support energy growth neutrality hypothesis. Energy consumption does not pay a pivotal role in economic growth and economic growth does not impact on energy consumption. In addition, the result also support EKC hypothesis where economic growth leads to environmental degradation in the short-run. Uganda should take care on its energy Table 2: ADF and PP unit root tests ADF test PP test Level 1st difference Level 1st difference Variables Test statistics Prob. Test statistics Prob. Adj. t-stat Prob. Adj. t-stat Prob. Ct 0.625201 0.9883 −4.904261 0.0004*** 0.595226 0.9874 −4.895838 0.0004*** ect −2.023753 0.2757 −5.528474 0.0001*** −2.021908 0.2764 −5.528490 0.0001*** gdpt −0.969973 0.7518 −4.534485 0.0011*** −0.880853 0.7813 −4.561942 0.0010*** *** indicates statistical significance at 1% Table 3: Optimal lag selection criteria Lag LogL LR FPE AIC SC HQ 0 55.52782 NA 6.77e-06 −3.388892 −3.250119 −3.343655 1 177.0059 211.6070* 4.80e−09* −10.64554* −10.09045* −10.46460* 2 184.5813 11.72953 5.36e−09 −10.55363 −9.582219 −10.23697 *indicates lag order selected by the criterion, LR: sequential modified LR test statistic (each test at 5% level). FPE: Final prediction error, AIC: Akaike information criterion, SC: Schwarz information criterion, HQ: Hannan-Quinn information criterion Table 4: Johansen Unrestricted Cointegration Rank Test (Maximum Eigenvalue and Trace) Hypothesized No. of CE (s) Eigenvalue Trace Max-Eigen Statistic Critical Value at 5% Prob. Statistic Critical Value at 5% Prob.** None * 0.386379 29.84717 29.79707 0.0493 15.13972 21.13162 0.2792 At most 1 0.291161 14.70745 15.49471 0.0655 10.66792 14.26460 0.1717 At most 2* 0.122174 4.039530 3.841466 0.0444 4.039530 3.841466 0.0444 Trace test indicates 1 cointegrating equation (s) at the 0.05 level, * denotes rejection of the hypothesis at the 0.05 level, **MacKinnon-Haug-Michelis (1999) P values Table 5: Long-run equilibrium Variables coefficient Standard errors t-statistic Ct 1.000000 ect −0.525794 0.55246 −0.95175 gdpt −1.853266 0.16542 −11.2034*** Constant 18.61247 Error correction term −0.292670 0.08426 −3.47344*** *** indicates statistical significance at 1% Otim, et al.: The Effects of Gross Domestic Product and Energy Consumption on Carbon Dioxide Emission in Uganda (1986-2018) International Journal of Energy Economics and Policy | Vol 12 • Issue 1 • 2022432 polices, the absence of causality between energy consumption and economic growth should not bring complacency because we see economic growth contributing to the environmental degradation in terms of increased CO2 emission per capita. We expect when Uganda starts to extract the oil in 2022 (Wolf and Potluri, 2018) the consumption of fossil fuels will increased, which is likely to have an adverse effect on the environment. The government expect to refine 150,000bbl daily and generates US$ 1billions in profits yearly through import substitution industrialization and export earnings (Wolf and Potluri, 2018). If we are to go by these figures, we expect growth in GDP per capita and energy consumption of fossil fuels, which may affect the environment adversely unless strong policy action are put in place to mitigate the possibility of environmental degradation. 4.5. VECM Estimation Diagnostics The estimated error correction term (ECT) is negative (−0.29267) and statistically significant at 1% confidence level (Tables 6 and 7). ECT indicates that any deviation from the long-run equilibrium between variables is corrected by about 29.27% for each period to return to the long-run equilibrium. In addition, figure 1 presents the plot of recursive estimate for Cumulative Sum (CUSUM) and Cumulative Sum Square (CUSUMSQ) of recursive residuals illustrated in figure 1. The coefficients were generated from VECM coefficients. The results indicate the absence of any instability of coefficients because the plots of CUSUM and CUSUMSQ statistics fall inside the critical band of the 5% confidence interval parameter stability. After estimating the VECM we tested for the robustness of the model through checking for serial correlation, residual normality, heteroskedasticity and model stability. Table 9 illustrates the results. We did not detect presence of serial correlation, and heteroskedasticity and the data is normality distributed. We can conclude that our model is pretty robust. 4.6. Variance Decomposition Variance decomposition in multivariate analysis used in economic forecasting (Lutkepohl, 2010). Tables 10-12 illustrate our result. Each row represents the percentage of the forecast error variance Table 6: Estimated coefficients Error Correction: Δct Δect Δgdpt CointEq1 −0.292670 [−3.47344]*** −0.035658 [−0.68657] 0.005443 [0.22625] Δct (−1) 0.129217 [0.75817] 0.018054 [0.17186] 0.047877 [0.98390] Δect (−1) −0.262820 [−0.82787] −0.042364 [−0.21650] 0.104407 [1.15187] Δgdpt (−1) −0.881842 [−1.31081] −0.442069 [−1.06608] 0.135236 [0.70407] Constant 0.058032 [2.46469] 0.003028 [0.20866] 0.025616 [3.81040] R2 0.331001 0.050923 0.129214 Adj. R2 0.228078 −0.095089 −0.004753 *** indicates statistical significance at 1% Table 7: Granger Causality test results Variables Short-run (or weak) Granger causality Long-run granger causality Δct Δect Δgdpt λi, i=1,2,3, Δct (−1) …………. 0.02954 (0.8635) 0.96805 (0.3252) −0.29267 (0.0018)*** Δect (−1) 0.68536 (0.4077) ………… 1.32681 (0.2494) 0.18749 (0.4984) Δgdpt (−1) 1.71821 (0.3083) 1.13652 (0.2864) ………. −0.01009 (0.8228) *** indicates statistical significance at 1% .The null hypothesis is that there is no causal relationship between variables. Values in the parentheses are P values for Wald tests with distribution. Δ is the first difference operator Table 9: VECM Post estimation test VEC residual serial correlation LM tests Residual normality tests VEC residual heteroskedasticity tests LM Stat Prob. Jarque-Bera Prob. Chi-sq Prob. 11.75350 0.2276 2.721445 0.25475 40.35525 0.7755 Table 10: Variance Decomposition of ct Period S.E. ct ect gdpt 1 0.072954 100.0000 0.000000 0.000000 2 0.093225 99.21564 0.327769 0.456591 3 0.102775 99.11120 0.297937 0.590868 4 0.110923 96.13280 1.456243 2.410956 5 0.119320 90.63697 3.958921 5.404113 6 0.128017 83.97681 7.144363 8.878828 7 0.137032 77.05142 10.52745 12.42112 8 0.146279 70.44342 13.80077 15.75581 9 0.155610 64.45631 16.79723 18.74646 10 0.164897 59.18318 19.45627 21.36055 Table 8: Pairwise granger causality tests Null hypothesis: F-statistic P-value Decision ect does not Granger Cause ct 2.12796 0.1554 Fail to reject the null hypothesis ct does not Granger Cause ect 0.07516 0.7859 Fail to reject the null hypothesis gdpt does not Granger Cause ct 7.36712 0.0111** Reject the null hypothesis ct does not Granger Cause gdpt 3.34088 0.0779 Fail to reject the null hypothesis gdpt does not Granger Cause ect 0.34235 0.5630 Fail to reject the null hypothesis ect does not Granger Cause gdpt 0.38798 0.5382 Fail to reject the null hypothesis *** indicates statistical significance at 1% Otim, et al.: The Effects of Gross Domestic Product and Energy Consumption on Carbon Dioxide Emission in Uganda (1986-2018) International Journal of Energy Economics and Policy | Vol 12 • Issue 1 • 2022 433 by the all the variables under investigation. We made a forecast of 10 year period. We divide our forecast in the short-run, mid-term and long-run. Table 10 shows that in the short-run 100% of the forecast error variance is explained by CO2 emission per capita itself. The contribution from energy consumption per capita, and GDP per capita to CO2 emission per capita are strongly exogenous implying that they have very weak influence in predicting CO2 emission per capita in the short-run. In the mid-term period, 90.6% CO2 emission per capita is explained by its standard innovation shock. This further still show that the weak influence of energy consumption per capita and GDP per capita is explained by CO2 emission per capita in Uganda. In the long-run, we see the influence of CO2 emission per capita on itself dwindling the further we move to the future. It implies that energy consumption and economic growth will start to have a strong endogenous influence on the carbon emission in Uganda. Table 11 shows that in the short-run 99.32% of the forecast error variance is explained by energy consumption itself and the contribution from GDP per capita and carbon emission per capita throughout the period of 10 years are very weak at less than 5%. In Table 12 we see only carbon emission per capita having fairly strong endogenous influence on GDP per capita while energy consumption is having weakly endogenous effect on GDP per capita. The forecast error variance reflect the Granger causality test results. 5. CONCLUSIONS This paper examines the long-run and causal links between emissions CO2, energy consumption, and economic growth in Uganda using VECM and Johansen cointegration test for long-run relationship and Granger causality models for 1986-2018 period. Empirical results suggest an evidence of a long-run relationship between variables at 5% levels of significance in Uganda. The estimated elasticity of CO2 emission per capita with respect to GDP capita is 1.856. The main results for the existence and direction of Granger causality are as follows: (i) The interesting finding is that energy consumption does not Granger cause CO2 emission per capita, although the main cause of CO2 emission in literature is energy consumption. Uganda can still continue to engage in oil exploration and start oil extraction to boost its economic growth without causing much damage to the environment at least in the short-run. (ii) There is a unidirectional causality flowing from GDP per capita to CO2 emissions per capita. The result does support the Environmental Kuznets Curve hypothesis, implying that economic growth leads to environmental pollution through CO2 emissions. (iii) There is no causal link between energy consumption per capita and GDP per capita. The result support growth neutrality hypothesis. The government of Uganda can pursue policies that promote energy consumption and economic growth such us increasing the level of industrialisation. To take care of environment green energy growth policy can be adopted for sustainable growth and development in the long-run. (iv) Long-run causality exists only for CO2 emission per capita equation. Figure 1: Plot of cusum and cusumsq recursive residuals Table 11: Variance Decomposition of ect Period S.E. ct ect gdpt 1 0.044967 0.671658 99.32834 0.000000 2 0.062794 0.381364 98.38387 1.234766 3 0.076261 0.650727 97.81901 1.530259 4 0.088252 0.859977 97.70996 1.430061 5 0.099174 0.995514 97.70586 1.298622 6 0.109195 1.109107 97.70500 1.185889 7 0.118496 1.209148 97.70150 1.089355 8 0.127215 1.295287 97.69797 1.006740 9 0.135443 1.369108 97.69441 0.936483 10 0.143247 1.432719 97.69057 0.876707 Table 12: Variance decomposition of gdpt Period S.E. ct ect gdpt 1 0.020829 20.20908 0.414469 79.37645 2 0.033550 27.24880 3.421960 69.32924 3 0.042056 29.14400 4.257522 66.59848 4 0.048961 29.03441 4.722091 66.24350 5 0.055195 28.41374 5.180753 66.40551 6 0.060931 27.74243 5.614048 66.64352 7 0.066245 27.11078 5.994692 66.89453 8 0.071213 26.53698 6.324101 67.13892 9 0.075894 26.02628 6.609572 67.36415 10 0.080331 25.57655 6.857346 67.56610 Cholesky Ordering ct, ect, gdpt Otim, et al.: The Effects of Gross Domestic Product and Energy Consumption on Carbon Dioxide Emission in Uganda (1986-2018) International Journal of Energy Economics and Policy | Vol 12 • Issue 1 • 2022434 The main result of this study indicates that GDP is key determinant of CO2 emissions in the long-run. As the growth in an economy is reflected through the increase in GDP at both nominal and real terms. This is because the growth in GDP is characterised by the growth in a number of economic activities besides energy consumption. A country like Uganda is predominantly agrarian and peasantry in nature. Therefore, the growth in its GDP is facilitated by unregulated economic practices. Therefore policies that are pro-growth may certainly conflict with policies that promote reduction in CO2 emission. The growth in GDP as the country strive to attain a middle income status is a precursor for future emission in Uganda. Thus, whereas growth in GDP is necessary for the welfare of the people, it should be sustainable for the posterity. The goals for economic growth and sustainable development are all important. Therefore, green growth strategy is likely to be the future of Uganda’s economy, if the country is to strike a balance between the two. The impact on energy consumption on the CO2 emission was found to be insignificant. This result suggests that there are other causal factors responsible CO2 emission. We used 33 years observations which may affect the predictive ability of our model. When more data are available in future, this study can be repeated to take care of the long-run time variation. However, the study did not consider the effect of population growth on the economic growth which could have been control variable to bite the influence of endogeneity in the model. Therefore, further research should investigate the impact of population activities on the environment as key factors for CO2 emission. Even though energy consumption is not the cause of CO2 emission in Uganda, we expect the long-run sustained economic growth to make energy consumption to be a key driver that will propel future emission like what is happening in China, where there is a huge utilisation of fossil fuels in construction and manufacturing sectors (Jiang et al., 2019; Yang et al., 2020; Zhu and Shan, 2020). The long-run scaled effect of economic growth and energy consumption on the environment can be mitigated by creating public awareness on the importance of green investments, supporting and adopting the clean use of energy technologies such as solar, hydro, geothermal and wind. REFERENCES Adom, P.K., Bekoe, W., Amuakwa-Mensah, F., Mensah, J.T., Botchway, E. (2012), Carbon dioxide emissions, economic growth, industrial structure, and technical efficiency: Empirical evidence from Ghana, Senegal, and Morocco on the causal dynamics. Energy, 47(1), 314-325. Akadiri, S.S., Bekun, F.V., Taheri, E., Akadiri, A.C. (2019), Carbon emissions, energy consumption and economic growth: A causality evidence. International Journal of Energy Technology and Policy, 15(2-3), 320-336. Alam, M.J., Begum, I.A., Buysse, J., Rahman, S., Van Huylenbroeck, G. (2011), Dynamic modeling of causal relationship between energy consumption, CO2 emissions and economic growth in India. Renewable and Sustainable Energy Reviews, 15(6), 3243-3251. Ang, J.B. (2007), CO2 emissions, energy consumption, and output in France. Energy Policy, 35(10), 4772-4778. Appiah, K., Du, J., Yeboah, M., Appiah, R. (2019), Causal relationship between industrialization, energy intensity, economic growth and carbon dioxide emissions: recent evidence from Uganda. International Journal of Energy Economics and Policy, 9(2), 237. Appiah, M.O. (2018), Investigating the multivariate Granger causality between energy consumption, economic growth and CO2 emissions in Ghana. Energy Policy, 112, 198-208. Bamwesigye, D., Kupec, P., Chekuimo, G., Pavlis, J., Asamoah, O., Darkwah, S.A., Hlaváčková, P. (2020), Charcoal and wood biomass utilization in Uganda: The socio economicand environmental dynamics and implications. Sustainability, 12(20), 8337. Beard, E., Marsden, J., Brown, J., Tombor, I., Stapleton, J., Michie, S., West, R. (2019), Understanding and using time series analyses in addiction research. Addiction, 114(10), 1866-1884. Bunn, D.W., Fezzi, C. (2008), A Vector Error Correction Model of the Interactions among Gas, Electricity and Carbon Prices: An Application to the Cases of Germany and the United Kingdom. Cheltenham: Markets for Carbon and Power Pricing in Europe: Theoretical Issues and Empirical Analyses. p145-159. Chang, C.C. (2010), A multivariate causality test of carbon dioxide emissions, energy consumption and economic growth in China. Applied Energy, 87(11), 3533-3537. Destek, M.A., Aslan, A. (2017), Renewable and non-renewable energy consumption and economic growth in emerging economies: Evidence from bootstrap panel causality. Renewable Energy, 111, 757-763. Dickey, D.A., Fuller, W. A. (1981), Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica: Journal of the Econometric Society, 1981, 1057-1072. Dinda, S. (2004), Environmental Kuznets curve hypothesis: A survey. Ecological Economics, 49(4), 431-455. Dogan, E. (2016), Analyzing the linkage between renewable and non- renewable energy consumption and economic growth by considering structural break in time-series data. Renewable Energy, 99, 1126- 1136. Engle, R.F., Granger, C.W. (1987), Co-integration and error correction: Representation, estimation, and testing. Econometrica: Journal of the Econometric Society, 1987, 251-276. Fan, W., Hao, Y. (2020), An empirical research on the relationship amongst renewable energy consumption, economic growth and foreign direct investment in China. Renewable Energy, 146, 598-609. GOU. (2015), Environment and Water Ministry: Water and Environment Sector Performance Report. Government of Uganda. Available from: http://www.mwe.go.ug Halicioglu, F. (2009), An econometric study of CO2 emissions, energy consumption, income and foreign trade in Turkey. Energy Policy, 37(3), 1156-1164. Hamilton, J.D. (2020), Time Series Analysis. United States: Princeton University Press. Held, D., Roger, C. (2018), Three models of global climate governance: From Kyoto to Paris and beyond. Global Policy, 9(4), 527-537. Horowitz, C.A. (2016), Paris agreement. International Legal Materials, 55(4), 740-755. Jiang, P., Yang, H., Ma, X. (2019), Coal production and consumption analysis, and forecasting of related carbon emission: Evidence from China. Carbon Management, 10(2), 189-208. Johansen, S. (1988), Statistical analysis of cointegration vectors. Journal of Economic Dynamics and Control, 12(2-3), 231-254. Kuznets, S. (1955), Economic growth and income inequality. The American Economic Review, 45, 1-28. Lütkepohl, H. (2010), Variance decomposition. In: Macroeconometrics and Time Series Analysis. London: Palgrave Macmillan. p369-371. Ma, X., Wang, C., Dong, B., Gu, G., Chen, R., Li, Y., Li, Q. (2019), Carbon emissions from energy consumption in China: Its measurement and driving factors. Science of the Total Environment, 648, 1411-1420. Markandya, A., Cabot-Venton, C., Beucher, O. (2015), Economic Assessment of the Impacts of Climate Change in Uganda. Otim, et al.: The Effects of Gross Domestic Product and Energy Consumption on Carbon Dioxide Emission in Uganda (1986-2018) International Journal of Energy Economics and Policy | Vol 12 • Issue 1 • 2022 435 United Kingdom: CDKN. Mawejje, J., Mawejje, D.N. (2016), Electricity consumption and sectoral output in Uganda: An empirical investigation. Journal of Economic Structures, 5(1), 21. Menegaki, A.N., Tugcu, C.T. (2016), The sensitivity of growth, conservation, feedback and neutrality hypotheses to sustainability accounting. Energy for Sustainable Development, 34, 77-87. Menyah, K., Wolde-Rufael, Y. (2010), Energy consumption, pollutant emissions and economic growth in South Africa. Energy Economics, 32(6), 1374-1382. Ozturk, I., Acaravci, A. (2010), CO2 emissions, energy consumption and economic growth in Turkey. Renewable and Sustainable Energy Reviews, 14(9), 3220-3225. Phillips, P.C., Perron, P. (1988), Testing for a unit root in time series regression. Biometrika, 75(2), 335-346. Protocol, K. (1997), United Nations framework convention on climate change. Kyoto Protocol, Kyoto, 19, 497. Rahman, M.M., Velayutham, E. (2020), Renewable and non-renewable energy consumption-economic growth nexus: new evidence from South Asia. Renewable Energy, 147, 399-408. ROU. (2020), Third National Development Plan (NDPIII) 2020/21- 2024/25: Uganda Vision 2040. The Republic of Uganda. Tamil Nadu: ROU. Sadik-Zada, E.R., Loewenstein, W. (2020), Drivers of CO2-emissions in fossil fuel abundant settings: (Pooled) mean group and nonparametric panel analyses. Energies, 13(15), 3956. Samuelson, R.D. (2014), The unexpected challenges of using energy intensity as a policy objective: Examining the debate over the APEC energy intensity goal. Energy Policy, 64, 373-381. Sekantsi, L.P., Okot, N. (2016), Electricity consumption-economic growth nexus in Uganda. Energy Sources, Part B: Economics, Planning, and Policy, 11(12), 1144-1149. Sghari, M.B.A., Hammami, S. (2016), Energy, pollution, and economic development in Tunisia. Energy Reports, 2, 35-39. Walters, D. (2021), Lumpy Social Goods in Energy Decarbonization: Why We Need More Than Just Markets for the Clean Energy Transition. Colorado: University of Colorado Law Review, Forthcoming. Wolf, S., Potluri, V.A. (2018), Uganda’s Oil: How Much, When, and How Will it be Governed? (No. 2018/179). WIDER Working Paper. Finland: World Institute for Development Economics Research Yang, J., Cai, W., Ma, M., Li, L., Liu, C., Ma, X., Chen, X. (2020), Driving forces of China’s CO2 emissions from energy consumption based on Kaya-LMDI methods. Science of the Total Environment, 711, 134569. Zhang, Y.J., Da, Y.B. (2015), The decomposition of energy-related carbon emission and its decoupling with economic growth in China. Renewable and Sustainable Energy Reviews, 41, 1255-1266. Zhu, B., Shan, H. (2020), Impacts of industrial structures reconstructing on carbon emission and energy consumption: A case of Beijing. Journal of Cleaner Production, 245, 118916.