TX_1~AT/TX_2~AT International Journal of Energy Economics and Policy | Vol 13 • Issue 3 • 2023102 International Journal of Energy Economics and Policy ISSN: 2146-4553 available at http: www.econjournals.com International Journal of Energy Economics and Policy, 2023, 13(3), 102-110. The Causal Relation between Energy Consumption, Carbon Dioxide Emissions, and Macroeconomic Variables in Somalia Zakarie Abdi Warsame1,3*, Maria Mohamed Ali1, Liban Bile Mohamud2, Farhia Hassan Mohamed1 1Faculty of Economics, SIMAD University, Mogadishu, Somalia, 2Somalia National Bureau of Statistics, Somalia, 3Kasmotute Insttitute for Technology, Logistics and Economics Research (KITLER), Somalia. *Email: zakariye1968@gmail.com Received: 20 January 2023 Accepted: 05 April 2023 DOI: https://doi.org/10.32479/ijeep.14262 ABSTRACT This study investigated the relationship between energy consumption, carbon dioxide emissions, and macroeconomic variables in Somalia with data spanning from 1990 to 2019 using ARDL model. The study found a negative long-run relationship between carbon dioxide emissions and energy consumption in Somalia, suggesting that improving access to clean energy can reduce the gradual rise of carbon dioxide emissions. The study also found that rising industrial value-added had a significant positive impact on energy consumption. Furthermore, findings from Cholesky’s variance decomposition showed that 13.13% of future fluctuations in energy consumption are due to shocks in carbon dioxide emission, 33.63% of future fluctuations in carbon dioxide emissions are due to shocks in energy consumption, 40.63% of future fluctuations in industrialization are due to shocks in energy consumption and 41.23% of future fluctuations in population are due to shocks in energy consumption. There was evidence of a bidirectional causality between: energy consumption and population. The study suggests adding renewable energy technologies to the energy portfolio. This would help reduce reliance on unstable energy sources and reduce the chance that changes in commodity prices will interrupt the energy supply, which eventually would help reduce the effects of climate change. Keywords: Energy Consumption, Industrialization, Trade Openness, Somalia and ARDL JEL Classifications: P18, O14, Q53 1. INTRODUCTION Energy consumption and related services to meet social and economic development and improve human health and welfare are increasing due to the requirement to meet basic human needs and productivity (Edenhofer et al., 2011). The energy development of a country is closely related to the economic development of the country. A country’s economic growth is directly influenced by its ability to provide energy to its citizens. Achieving the Millennium Development Goals (MDGs) depends heavily on access to energy. It is undoubtedly evident that energy inadequacies have a close association with poverty indicators, such as illiteracy, life expectancy, infant mortality, fertility rates, and rapid urbanization in developing countries like Somalia; this is because rural residents migrate to urban areas in search of better living conditions and social amenities (Lipton and Ravallion, 1995). Due to the limited energy supply in Somalia, the rapid growth of the urban population is currently being hampered by energy insecurity. Around 80-90% of all energy consumption in Somalia comes from wood and charcoal (African Development Bank, 2015). Electricity provision in Somalia has been the primary responsibility of the country’s thriving private sector since the collapse of the central government in 1991. As of right now, the total production capacity is around 106 Megawatts. While most utilities still use gasoline power plants to generate electricity, hybrid systems that take advantage of renewable sources like solar and wind are attracting more and more attention and funding. Recent research by the African Development Bank found that Somalia has the greatest resource potential for coastal wind power in Africa, with the capacity to produce between 30,000 and 45,000 Megawatts. Solar panels could generate more than 2,000 kWh/m2 of energy. This Journal is licensed under a Creative Commons Attribution 4.0 International License Warsame, et al.: The Causal Relation between Energy Consumption, Carbon Dioxide Emissions, and Macroeconomic Variables in Somalia International Journal of Energy Economics and Policy | Vol 13 • Issue 3 • 2023 103 Around one-sixteenth of the population, by some estimates, has access to modern power. Somalia has more expensive taxes than its neighbors, Kenya and Ethiopia (U.S. Agency for International Development). Somalia consumed 12,100,621,000 BTU (0.001 quadrillion BTU) of energy in 2017, representing 0.00% of global energy consumption. The country produced 156,621,000 BTU (0.00 quadrillion BTU), covering 1% of its annual energy consumption (Somalia Energy Statistics-Worldometer, n.d.). Despite this, since 1960, carbon dioxide emissions from the combustion of fossil fuels have tripled. Concerns have grown due to the realization that human-caused carbon dioxide emissions significantly contribute to climate change. Emissions of carbon dioxide and models of economic performance may be critical for understanding the connections between population growth and financial performance. Rapid economic growth and demographic expansion contribute to environmental deterioration (Chandia et al., 2018). Several economic and legal factors are making the environmental situation around the world worse. These factors operate in different areas and have varying degrees of impact and consequences. They include a macroeconomic policy that encourages the overuse of natural resources, an investment policy that prioritizes using natural resources, and a sectoral policy that needs improving, especially in the fuel and energy industry (Shpak et al., 2022). Somalia’s CO2 emissions are relatively low compared to other countries due to its limited industrialization and low per capita energy consumption. Somalia’s energy sector is primarily based on fossil fuels, with oil accounting for about 95% of the country’s total energy consumption. However, due to the ongoing civil conflict and lack of infrastructure, energy access is limited, and the country’s total CO2 emissions remain low. According to the (World Bank, n.d.), Somalia’s CO2 emissions in 2018 were estimated to be 0.06 metric tons per capita, which is significantly lower than the global average of 4.8 metric tons per capita. However, it is important to note that Somalia, like other developing countries, is vulnerable to the impacts of climate change, including sea-level rise, droughts, and floods. Besides, Conservation policies can greatly affect how well the economy does because every economy depends a lot on how much energy is used. So, testing macroeconomic and environmental variables in the real world is important, as it is crucial to clarifying policy implications and recommendations. Since per capita income is associated with energy consumption, economic growth can also be identified as the primary cause behind the increase in energy consumption over the last decade (Asumadu-Sarkodie and Owusu, 2016a). Nevertheless, no consensus has been achieved about the pattern of the causal link between rising macroeconomic output and energy consumption in Somalia. In light of this, the study investigates the relationship between macroeconomic variables, carbon dioxide emissions, and energy consumption in Somalia. The research makes an effort to address the gap in the literature on energy-emissions economic analysis, which has been spotty and scarce in Somalia. In order to assess how each random innovation affects energy usage, carbon dioxide emissions, and increased macroeconomic output the research estimates the variance decomposition using Cholesky’s approach. Finally, reliable estimation methods based on Auto Regressive Distributive Lag ARDL, and granger causality test are used to provide more extensive scheme suggestions from Somalia’s energy consumption. The paper is structured as follows: Section 2 reviews the relevant literature, Section 3 describes the research methodology and data, Section 4 presents the study’s findings, and Section 5 concludes with policy implications. 2. LITERATURE REVIEW The literature provides ample documentation of the dynamic causal connection between energy consumption and environmental pollution. The primary objective of these investigations was to describe temporal relationships, but bivariate models were used extensively. There appears to be no consensus regarding the dynamic causal relationship between energy consumption and environmental pollution. Possible causes of inconclusive results include misspecification of estimated models, bias from omitted variables, or failure to select true lag lengths (which are very sensitive to Granger causality). Table 1 represents a few of the existing literature examined in the study with their subsequent econometric method, the length of the data employed, and the findings of their study. 3. METHODOLOGY The study examines the causal nexus between energy consumption, carbon dioxide and macroeconomic variables in Somalia. A time series data spanning from 1990 to 2019 were employed from World Bank, SESRIC and World data. Six study variables were used in the study which include EC-energy-consumption (kilogram of oil equivalent per capita); CO2-carbon dioxide emissions (kt); TO- Trade Openness; IND-industry, value added (Constant 2015 US$), which is a proxy for industrialization; GDPPC-RGDP per capita (Constant 2015 US$); and POP-population. A linear representation of the relationship between energy consumption, carbon dioxide and macroeconomic variables in Somalia is showed in Eq. (1): InEC F InCO InTO InIND InPOP InGDPPCt t t t t t= ( , ), , ,2 (1) Where InECt, InCO2t, InTOt, InINDt, lnPOPt and lnGDPPCt represent a natural logarithmic transformation of CO2, TO, IND, POP and GDPPC for a more stable data variance. The empirical specifications for the model can be quantified as: Warsame, et al.: The Causal Relation between Energy Consumption, Carbon Dioxide Emissions, and Macroeconomic Variables in Somalia International Journal of Energy Economics and Policy | Vol 13 • Issue 3 • 2023104 InEC InCO InTO InIND InPOP InGDPC t t t t t t t � � � � � � � � � � � � � � 0 1 2 3 4 5 2 (2) Where InECt is the dependent variable, while InCO2t, InTOt, InINDt, InPOPt and InGDPPCt are the explanatory variables in year t, εt is the error term, and β0, β1, β2, β3, β4 and β5 are the elasticities to be estimated. The first step in testing for cointegration is investigating the order of integration of the variables. We apply Augmented Dickey-Fuller (ADF) and Philips-Perron (PP) stationarity tests to determine if the series are cointegrated. Once the series become stationary, we select the lag order and investigate whether there are cointegrating relationships between variables. To examine the long run relationship among the model variables, there are several tests of cointegration. The first one that has been extensively used and discussed in the literature is the popular Engle and Granger test which is applicable only for same order integrated variables. Subsequently, many other approaches have been developed some of which are the Error correction Cointegration technique of Johansen which is more general and flexible than the Engle and Granger approach, Phillips and Ouliaris test, Johansen and Juselius test, the Structural Error Correction Model (ECM) proposed by (Boswijk, 1994), and the test suggested by (Banerjee et al., 1998a) which is based on the t-test for the null hypothesis. However, these standard approaches have been criticized as being highly unreliable in small samples, inconsistent with different order integrated variables, lead to significantly misleading results and biased against the rejection of null hypothesis (no-cointegration) which requires an adjustment for critical values (Shahbaz et al., 2015). Hence, in order to increase the power of test, more robust cointegration technique is employed which is autoregressive distributed lag (ARDL) bounds testing approach. Following the empirical work of (Sarkodie and Adams, 2018) the ARDL cointegration equation can be written as: � � � � � � � � � � � � InEC InCO InTO InIND InPOP In t t t t t � � � � � � 0 1 1 2 1 3 1 4 1 5 2 GGDPPC InEC InCO InTO t i q t k i p t k i p t k � � � � � � � � � � � � � � � � 1 0 1 0 2 0 32 � � � �� � � � � � � � � � � � � � � � � i p t k i o p t k i p t k t InIND InPOP InGDPPC 0 4 5 0 6 � � � � kk (3) Where α0 is the constant, α1–α6 are the coefficient of the short- tun variables, β1-β5 are the elasticities of long-run parameters, q indicates the explained’s optimal lags, p shows the optimal lags of the explanators, Δ is the first difference sign showing short run variables and εt is the error term. The ARDL cointegration approach begins with bound testing, which is then regressed using Ordinary Least Squares (OLS). The null hypothesis (H0): β1 = β2 = β3 = β4 = β5 = β6 = β7 = β8= 0 implies variables are not cointegrated in the long-run whereas the alternative hypothesis (H1): β1 ≠ β2 ≠ β3 ≠ β4 ≠ β5 ≠ β6 ≠ β7 ≠ 0 implies variables are cointegrated in the long-run. The Wald-F statistics and critical values were employed to test the null hypothesis. If the Wald-F statistics exceed the upper bound critical values, the null hypothesis is rejected, indicating that the variables are linked in the long run and vice versa. 4. RESULTS AND DISCUSSION 4.1. Descriptive Analysis and Correlation Matrix We examined the characteristics of the data series using descriptive statistics presented in Table 2. Results in Table 2 report the mean of Energy Consumption (2.48), Carbon dioxide (12.79), Trade Table 1: Summary of the related literature Reference Time period Econometric approach DV IV Findings (Asumadu-Sarkodie and Owusu, 2017) 1960-2013 VECM EUSE CO2, FID, IND, GDPPC, POP EC↔FD, EC↔IND, EC↔POP, CO2↔FD, CO2↔GDPPC (Warsame, 2023) 1990-2019 ARDL CO2 FDI, REC, GDPPC, PG, K CO2→PG, REC→PG (Shahbaz et al., 2015) 1980-2012 VECM CO2 EC, FP, TSVA CO2↔EC, TSVA↔CO2, FP→CO2, FP→EC, FP→TSVA (Rafindadi and Ozturk, 2015) 1971-2012 ARDL NGC GDP, LF GDP does not Granger-cause NGC (Warsame and Sarkodie, 2021) 1985-2017 NARDL DEFO EC, RGDPC, and PG PG↔GDP, GDP→PG, GDP→EC (Warsame et al., 2022) 1990-2017 ARDL ED REC, POP. IQ, RGDPC, and K No causality is observed from renewable energy to environmental degradation and vice versa. (Lin et al., 2015) 1980-2011 VECM CO2 GDP, EC, and POP Weak long-run causality from EC to CO2 (Asumadu-Sarkodie and Owusu, 2016b) 1980-2012 VECM CO2 GDP, EC, and POP CO2↔EC, GDP↔EC, POP→CO2 (Cerdeira Bento and Moutinho, 2016) 1960-2011 ARDL CO2 GDP, REEP, NREEP, and INT GDP→REEP, NREEP→REEP (Mohiuddin et al., 2016) 1971-2013 VECM CO2 EC and GDP EC→CO2 (Chen et al., 2019) 1980-2014 ARDL CO2 GDP, R, N, and T long-run causality from GDP, the square of GDP, REC, N, and T to CO2. (Salahuddin et al., 2015) 1980-2012 FMOLS CO2 GDP, EC, and FD Long-run causality from GDP to CO2 (Dong et al., 2018) 1993-2016 VECM CO2 GDP, FF, NU, and RE Long-run causality among CO2, FF, NU, and REC Warsame, et al.: The Causal Relation between Energy Consumption, Carbon Dioxide Emissions, and Macroeconomic Variables in Somalia International Journal of Energy Economics and Policy | Vol 13 • Issue 3 • 2023 105 Openness (1.44), Industrialization (7.86), Population (6.57), and GDP per capita (2.47). Besides, Industrialization and Population have the highest maximum values of 8.27, and 6.85 respectively. Energy Consumption, Population, and GDDPC are positively skewed while CO2, Trade Openness, and Industrialization are negatively skewed. Furthermore, the correlation of the sampled variables presented in Table 3 shows that Carbon dioxide, Trade Openness, Industrialization, Population, and GDP Per Capita are negatively correlate with Energy Consumption in Somalia. 4.2. Unit Root Test Testing the unit root properties is a prerequisite in time series modeling, specifically ARDL. Hence, Augmented Dickey-Fuller (ADF) and Philips perron (PP) tests were utilized to circumvent spurious regression results. The unit root analysis reported in Table 4 shows lnEC is stationary at level [I (0)], whereas the remaining series has unit root. However, Table 4 shows that most of the series are integrated at first difference [I (1)] while only lnEC is integrated at [I (0)]. Since none of the variables are stationary at second difference I (2), we proceeded to estimate the bounds test cointegration. 4.3. Cointegration Test Results of the bounds test presented in Table 5 examine the presence of long-run co-integration between energy consumption and the explanator variables. However, the results show the Wald F-statistics (13.06530) is above the upper bound critical value (3.38) at 5% significance level. This infers the variables are cointegrated in the long run. 4.4. ARDL Long-run and Short-run Results with Diagnostics The long-run estimations of the ARDL method are presented in Table 6, with some diagnostic test statistics. The results show a negative long-run significant relationship between carbon dioxide emission and energy consumption in Somalia. In other words, carbon dioxide emissions are negatively related to energy consumption in the long run, which has a policy implication in Somalia. It is likely that reducing carbon dioxide emissions through the adoption of renewable and clean energy technologies will eventually improve energy consumption in Somalia in the long run. This finding is in line with the findings of (Asumadu- Sarkodie and Owusu, 2017). The impact of rising industrial value added also has significant positive impact on energy consumption. The rise in industrial activities requires more energy to contribute to the gross domestic product. A 1% rise in industrial value-added increases energy consumption by 1.34% in the long run. Population is found to be negatively related to energy consumption in the long run. This implies that a 1% increase in population leads a decrease of 2.040% to energy consumption in the long run. This contradictory finding may be attributable to the fact that a large proportion of the Somali population resides in rural areas and is typically unable to obtain oil or fuel. Finally, the study found that variables of trade openness and GDPPC are insignificant in the long run. Our results indicating that an increase in industrial value-added leads to increased energy consumption are in line with the findings of (Lovins, 1990), (Shahbaz and Lean, 2011). Table 2: Descriptive statistics of variables Stats LEC LCO2 LTO LIND LPOP LGDPPC Mean 2.483004 12.785123 1.435547 7.862481 6.571825 2.474741 Median 2.483861 2.795866 1.411702 7.910259 6.566554 2.461671 Maximum 2.764392 2.863323 2.027635 8.265808 6.852656 2.622711 Minimum 2.357120 2.690196 0.750508 7.434153 6.341225 2.340999 SD 0.076813 0.043834 0.463578 0.272687 0.165930 0.089644 Skewness 1.392708 −0.514309 −0.011466 −0.143373 0.179796 0.129722 Jarque-Bera 31.04760 1.446313 3.062268 2.508708 2.246811 1.833647 P-value 0.000000 0.485218 0.216290 0.285260 0.325171 0.399787 Table 3: Correlation matrix LEC LCO2 LTO LIND LPOP LGDPPC LEC 1 LCO2 −0.0042 1 LTO −0.7191 0.3411 1 LIND −0.7721 0.2503 0.9797 1 LPOP −0.8085 0.2914 0.9754 0.9821 1 LGDPPC −0.6198 0.5987 0.9138 −0.8959 0.8948 1 Table 4: Unit root Variables T-statistics at level ADF PP lnEC −7.4709*** −6.2445*** lnCO2 −2.6578 −2.3435 lnTO −1.1355 −2.9312 lnIND −2.2516 −2.2305 lnPOP −2.9307 −2.5392* lnGDPPC −2.3044 −2.5738 At first difference ΔInCO2 −3.4250* −3.4618* ΔInTO −4.5596*** −4.5596*** ΔInIND −5.9256*** −5.8817*** ΔInPOP −4.4759*** −1.6335 ΔInGDPPC −5.6156*** −5.6063*** ***, **, *Indicate the significance level at 1%, 5%, and 10%. Δ denotes first difference operator. The T-statistics reported are the intercept and trend Table 5: F bounds test F-statistic Level of significance (%) Bounds test critical values I (0) I (1) 13.06530 1 3.06 4.15 5 2.39 3.38 10 2.08 3 Warsame, et al.: The Causal Relation between Energy Consumption, Carbon Dioxide Emissions, and Macroeconomic Variables in Somalia International Journal of Energy Economics and Policy | Vol 13 • Issue 3 • 2023106 Our results indicating the negative effect of population increase on energy consumption are corroborated by numerous studies such as (Kunvitaya and Dhakal, 2017), (Ohlan, 2015), and (Otsuka, 2018) who conclude that population density has a negative impact on energy consumption. Diagnostic check results show no serial correlation, heteroscedasticity, model misspecification, and normality problems in the ARDL model. Also, the coefficients of the ARDL model are found stable over the sample period according to the CUSUM and CUSUM-square tests, and test results are presented in Figures 1 and 2, respectively. The short-run elasticities are computed as the estimated coefficients of the first differenced variables. The short-run results are reported in Table 7. Carbon dioxide exerts negative impact on energy consumption marginally. In short-run, energy consumption will decline by 1.4% due to a 1% increase in carbon dioxide emission. The effect of trade openness on energy consumption is negative and highly significant. This indicate that a 1% increase in trade openness leads a decrease of 0.39% to the energy consumption in the short run. The economic activities in industrial sector are positively associated with energy consumption. It is found that 1% increase in industrial value added will cause 0.46% energy consumption rise. Findings also revealed that Population rise is positively affects energy consumption as 1% increase in population leads an increase of 47% to the energy consumption in the short run. The impact of economic growth on energy consumption is positive and highly significant. A 1% rise in economic growth will increase energy consumption by 1.04%. The significance of error correction term implies that change in the response variable is a function of disequilibrium in the cointegrating relationship and the changes in other explanatory variables. The coefficient of ECTt-1 shows speed of adjustment from short-run to long-run and it is statistically significant with negative sign. (Banerjee et al., 1998b) noted that significant lagged error term with negative sign is a way to prove that the established long-run relationship is stable. The deviation of energy consumption from short-run to the long-run is corrected by 12.65% each year. Table 6: Long run results and diagnostics Variables Coefficient C 51.3108 (3.3227)** lnCO2 ‒5.2763 (‒1.9771)* lnTO ‒0.1054 (‒1.2831) lnIND 1.3428 (2.2224)* lnPOP 2.0407 (‒2.7432)** lnGDPPC ‒0.3539 (‒0.6727) Reset test 0.6933 (0.5105) Serial correlation 2.6739 (0.1478) Heteroskedasticity 20.6589 (0.6927) Normality 0.7342 (0.6927) ***, **,* Indicate significance levels at 1%, 5%, and 10%. The T-statistics are reported in (..), p-values are in [..] Table 7: Short run ECM results Variables Coefficient ΔINECt-1 −0.69332 (−7.8215) *** ΔInCO2 −1.429 (−8.8308) *** ΔInCO2t-1 1.2672 (−8.8308) *** ΔInTO −0.3945 (−9.6429) *** ΔInTOt-1 −0.2419 (−5.8585) *** ΔInIND 0.4623 (9.3141) *** ΔInPOP 47.8542 (6.9768) *** ΔInPOPt-1 −106.015 (−8.4944) *** ΔInPOPt-2 29.5116 (7.3079) *** ΔInGDPPC 1.0412 (−6.9478) *** ΔInGDPPCt-1 1.0603 (7.8346) *** ΔInGDPPCt-2 0.6323 (6.5418) *** ECTt-1 −0.6583 (−12.6511) *** ***, **,* Indicate significance levels at 1%, 5%, and 10%. The T-statistics are reported in (..), P values are in [..] Figure 1: Assessing parameter stability using CUSUM test Figure 2: Assessing parameter stability using CUSUM square test Warsame, et al.: The Causal Relation between Energy Consumption, Carbon Dioxide Emissions, and Macroeconomic Variables in Somalia International Journal of Energy Economics and Policy | Vol 13 • Issue 3 • 2023 107 4.5. Granger Causality Test In order to determine the direction of causality between variables, we conducted the Granger causality test shown in Table 8. From CO2 to energy consumption, trade openness to CO2, population to CO2, GDPPC to industrialization, and GDPPC to population we observed unidirectional causation. There are bidirectional causal relationships between population and energy consumption, industrialization and CO2, industrialization and trade openness, population and trade openness, GDPPC and trade openness, and population and industrialization. 4.6. Variance Decomposition This section estimates the response of variables to random innovation affecting the variables in the VAR using the Cholesky’s method of variance decomposition. Evidence from Table 9 A shows that 13.13% of future fluctuations in lnEC are due to shocks in lnCO2, 11.43% of future fluctuations in lnEC are due to shocks in lnTO, 5.46% of future fluctuations in lnEC are due to shocks in lnPOP, 3.92% of future fluctuations in lnEC are due to shocks in lnIND, and 0.99% of future fluctuations in lnEC are due to shocks lnGDPPC. As a policy implication for Somalia, carbon dioxide emissions affect energy consumption in the future more than trade openness population, industrialization, and GDP per capita. Evidence from Table 9 B shows that almost 33.63% of future fluctuations in lnCO2 are due to shocks in lnEC, 18.73% of future fluctuations in lnCO2 are due to shocks in lnTO, 16.3% of future fluctuations in lnCO2 are due to shocks in lnPOP, 3.36% Table 8: Pairwise granger causality Null hypothesis Obs F-Statistic Prob. InCO2 does not granger cause InEC 28 3.69957 0.0405 InEC does not granger cause InCO2 28 1.22018 0.3136 InTO does not granger cause InEC 28 2.04205 0.1526 InEC does not granger cause InTO 28 1.17761 0.3259 InIND does not granger cause InEC 28 1.0382 0.3701 InEC does not granger cause InIND 28 2.03355 0.1537 InPOP does not granger cause InEC 28 2.56318 0.0989 InEC does not granger cause InPOP 28 8.37181 0.0019 InGDPPC does not granger cause InEC 28 0.0409 0.9600 InEC does not granger cause InGDPPC 28 2.40318 0.1128 InTO does not granger cause InCO2 28 5.47321 0.0114 InCO2 does not Granger Cause InTO 28 2.13027 0.1417 InIND does not granger cause InCO2 28 5.33058 0.0125 InCO2 does not granger cause InIND 28 3.47038 0.0482 InPOP does not granger cause InCO2 28 11.7016 0.0003 InCO2 does not granger cause InPOP 28 1.1926 0.3215 InGDPPC does not granger cause InCO2 28 1.90639 0.1714 InCO2 does not granger cause InGDPPC 28 0.25898 0.7741 InIND does not granger cause InTO 28 2.98343 0.0705 InTO does not granger cause InIND 28 4.64893 0.0202 InPOP does not granger cause InTO 28 3.75578 0.0388 InTO does not granger cause InPOP 28 3.04342 0.0672 InGDPPC does not granger cause InTO 28 6.79864 0.0048 InTO does not granger cause lnGDPPC 28 3.48718 0.0476 InPOP does not granger cause InIND 28 15.0648 0.0000 InIND does not granger cause InPOP 28 11.6456 0.0003 InGDPPC does not granger cause InIND 28 11.7177 0.0003 InIND does not granger cause InGDPPC 28 0.40083 0.6744 InGDPPC does not granger cause InPOP 28 19.0959 0.0000 InPOP does not granger cause InGDPPC 28 0.44707 0.6449 Table 9: Variance decomposition A) Variance decomposition of lnEC Period S.E. INEC INCO2 INTO ININD INPOP INGDPPC 1 0.037776 100 0 0 0 0 0 2 0.047324 81.74989 7.120279 5.224611 0.222687 5.273605 0.408927 3 0.0576 73.40587 13.42851 6.978587 1.376377 4.473519 0.337144 4 0.061162 71.62573 12.35961 9.294517 2.411826 3.967663 0.340651 5 0.06209 69.58832 12.11219 11.04883 2.903981 4.01573 0.330948 6 0.062742 68.16538 12.45138 10.86428 3.179632 4.819499 0.519825 7 0.063678 66.25832 13.11106 11.0774 3.383104 5.332493 0.83762 8 0.064282 65.38054 13.30776 11.45954 3.603291 5.278576 0.970297 9 0.064652 65.20435 13.20886 11.5053 3.827624 5.28394 0.969932 10 0.064878 65.07637 13.12598 11.42653 3.922963 5.459732 0.988424 B) Variance decomposition of lnCO2 Period S.E. lnEC lnCO2 lnTO lnIND lnPOP lnGDPPC 1 0.018160 18.98679 81.01321 0.000000 0.000000 0.000000 0.000000 2 0.020924 14.54022 61.0513 0.217973 0.549562 22.00749 1.633451 3 0.028213 32.85115 42.86229 3.161401 0.315152 19.37486 1.435139 4 0.035320 39.48501 32.50462 7.579748 0.598448 17.79076 2.041415 5 0.039651 36.96935 29.0155 15.33211 0.614804 15.64264 2.425601 6 0.041208 35.52213 27.68793 19.21259 0.737222 14.48418 2.355943 7 0.041704 35.06113 27.03512 19.77435 0.837832 14.93328 2.358295 8 0.042176 34.5351 26.63378 19.3365 1.069199 15.7669 2.658517 9 0.042800 34.04188 26.10001 18.98674 1.554874 16.23534 3.081156 10 0.043420 33.62897 25.68895 18.72736 2.298263 16.29576 3.360701 C) Variance decomposition of lnTO Period S.E. lnEC lnCO2 lnTO lnIND lnPOP lnGDPPC 1 0.060317 0.097471 14.91304 84.98949 0.000000 0.000000 0.000000 2 0.085146 0.368801 9.412168 66.19750 13.24972 5.954974 4.816833 3 0.115074 7.745887 9.552816 38.76711 22.19934 13.21844 8.516404 (Contd...) Warsame, et al.: The Causal Relation between Energy Consumption, Carbon Dioxide Emissions, and Macroeconomic Variables in Somalia International Journal of Energy Economics and Policy | Vol 13 • Issue 3 • 2023108 of future fluctuations in lnCO2 are due to shocks in lnGDPPC, and 2.3% of future fluctuations in lnCO2 are due to shocks in lnIND. As a policy implication for Somalia, energy consumption affects carbon dioxide emissions in the future more than trade openness, GDP per capita, industrialization, and population. Moreover, evidence from Table 9 C shows that, almost 32.88% of future fluctuations in lnTO are due to shocks in lnIND, 23.33% of future fluctuations in lnTO are due to shocks in lnCO2, 15.14% of future fluctuations in lnTO are due to shocks in lnEC, 8.8% of future fluctuations in lnTO are due to shocks in lnPOP, and 7.02% of future fluctuations in lnTO are due to shocks in lnGDPPC. As a policy implication for Somalia, industrialization affects trade openness in the future more than carbon dioxide emission, energy consumption, population, and GDP per capita. Evidence from Table 9 D shows that almost 40.63% of future fluctuations in lnIND are due to shocks in lnEC, 25.21% of future fluctuations in lnIND are due to shocks in lnTO, 7.74% of future fluctuations in lnIND are due to shocks in lnCO2, 6.97% of future fluctuations in lnIND are due to shocks in lnPOP, and 0.92% of future fluctuations in lnIND are due to shocks lnGDPPC. As a policy implication for Somalia, energy consumption affects industrial value added in the future more than trade openness, carbon dioxide emissions, population and GDP per capita. In addition, evidence from Table 9 E shows that almost 41.23% of future fluctuations in lnPOP are due to shocks in lnEC, 28.24% of future fluctuations in lnPOP are due to shocks in lnTO, 4.84% of future fluctuations in lnPOP are due to shocks in lnIND, 4.69% of future fluctuations in lnPOP are due to shocks in lnCO2, and 2.03% of future fluctuations in lnPOP are due to shocks in lnGDPPC. As a policy implication for Somalia, energy consumption Table 9: (Continued) C) Variance decomposition of lnTO Period S.E. lnEC lnCO2 lnTO lnIND lnPOP lnGDPPC 4 0.144888 10.07868 13.33237 25.56425 28.80558 12.84684 9.372283 5 0.173012 12.94577 14.73421 18.10619 33.18196 11.59211 9.439766 6 0.199343 15.71125 16.61490 1,364,181 34.92462 10.33313 8.774288 7 0.217774 16.50086 19.00609 11.50073 35.44629 9.461863 8.084161 8 0.228622 16.27145 21.09667 10.77755 35.15938 9.075692 7.619251 9 0.234828 15.65524 22.70668 11.27476 34.13013 8.942591 7.290592 10 0.239247 15.13880 23.32543 12.83254 32.88161 8.797875 7.023740 D) Variance decomposition of lnIND Period S.E. lnEC lnCO2 lnTO lnIND lnPOP lnGDPPC 1 0.053802 51.95368 4.218534 11.05033 32.77745 0.000000 0.000000 2 0.060476 44.20283 12.19904 15.10017 26.56296 1.932907 0.002093 3 0.064266 39.18167 14.51023 14.33457 25.18027 5.842866 0.960392 4 0.070585 32.61441 16.41487 18.64761 20.87722 9.686311 1.759578 5 0.073798 29.86625 16.04012 23.38977 19.10734 9.757484 1.839036 6 0.076304 28.64037 15.04378 26.54260 18.29666 9.735596 1.740986 7 0.080212 29.57774 13.70947 27.97173 17.74547 9.406054 1.589534 8 0.087632 33.72539 11.63210 26.98443 17.65156 8.635057 1.371463 9 0.097614 38.05545 9.435389 25.74610 17.89929 7.743952 1.119817 10 0.107793 40.63962 7.738617 25.20531 18.52484 6.970276 0.921331 E) Variance decomposition of lnPOP Period S.E. lnEC lnCO2 lnTO lnIND lnPOP lnGDPPC 1 0.000978 29.95410 1.919819 0.520289 5.368368 62.23743 0.000000 2 0.002540 30.70284 2.044221 2.673434 4.960494 58.34512 1.253891 3 0.004913 30.93122 3.410568 10.14105 4.050014 48.75990 2.707250 4 0.007649 30.04796 4.415351 17.55936 3.410027 41.14973 3.417574 5 0.010448 30.31854 4.905000 22.67687 3.106397 35.39828 3.594918 6 0.013174 31.73386 5.126825 25.85679 3.013320 30.82603 3.443170 7 0.015830 33.94238 5.233862 27.49914 3.148990 27.06262 3.113004 8 0.018444 36.54531 5.213597 28.14391 3.512493 23.86276 2.721928 9 0.021002 39.06284 5.026007 28.29266 4.093843 21.17509 2.349561 10 0.023450 41.23321 4.689210 28.24065 4.843619 18.96360 2.029716 F) Variance decomposition of lnGDPPC Period S.E. lnEC lnCO2 lnTO lnIND lnPOP lnGDPPC 1 0.028245 4.146526 1.049299 9.764113 0.530419 63.62860 20.88105 2 0.042112 4.609519 9.749053 36.46300 4.197690 32.40336 12.57737 3 0.043942 4.401092 11.99666 37.00492 4.043623 30.58769 11.96601 4 0.045385 6.503746 11.98368 37.32893 4.278083 28.67907 1122649 5 0.047393 9.273601 12.79228 36.77025 4.248973 26.59928 10.31562 6 0.050663 15.24251 12.42754 34.66226 5.191256 23.27877 9.197672 7 0.05405 19.39081 10.93782 33.07414 7.875717 20.45694 8.264568 8 0.056983 20.93201 10.10655 32.07623 10.71740 18.49593 7.671884 9 0.059722 21.76906 10.19500 30.18350 13.45841 17.06708 7.326937 10 0.062256 22.65140 10.90741 27.98759 15.46832 15.91398 7.071297 Warsame, et al.: The Causal Relation between Energy Consumption, Carbon Dioxide Emissions, and Macroeconomic Variables in Somalia International Journal of Energy Economics and Policy | Vol 13 • Issue 3 • 2023 109 affects industrialization in the future more than trade openness, industrialization, carbon dioxide emissions, and GDP per capita. Finally, evidence from Table 9 F shows that almost 27.99% of future fluctuations in lnGDPPC are due to shocks in lnTO, 22.65% of future fluctuations in lnGDPPC are due to shocks in lnEC, 15.91% of future fluctuations in lnGDPPC are due to shocks in lnPOP, 15.47% of future fluctuations in lnGDPPC are due to shocks in lnIND, and 10.91% of future fluctuations in lnGDPPC are due to shocks in lnCO2. As a policy implication for Somalia, trade openness affects population in the future more than energy consumption, population, industrialization and carbon dioxide emissions. 5. CONCLUSION AND POLICY IMPLICATIONS The study examined the causal relationship between energy consumption, carbon dioxide emissions, and macroeconomic variables in Somalia with a data spanning from 1990 to 2019 using the ARDL model. The summary of findings are as follows: The results show a negative long-run significant relationship between carbon dioxide emission and energy consumption in Somalia. We can say that energy policies aimed at improving access to clean energy consumption in Somalia will reduce the gradual rise of carbon dioxide emission in Somalia. The impact of rising industrial value added also has significant positive impact on energy consumption. The rise in industrial activities requires more energy to contribute to the gross domestic product. A 1% rise in industrial value-added increases energy consumption by 1.34% in the long run. Population is found to have a negative long-term relationship with energy consumption. In the long term, a 1% increase in population results in a 2.040% decrease in energy consumption. This contradictory finding may be attributable to the fact that a large proportion of the Somali population resides in rural areas and is typically unable to obtain oil or fuel. Further, the study found that variables of trade openness and GDPPC are insignificant in the long run. Carbon dioxide emission and trade openness are adversely affected to the energy consumption in the short run. Contrary to this, industrialization, population and GDDPC increase energy consumption in Somalia. ECM’s significant error coefficient confirms the existence of a long-run relationship between the variables. With regards to Granger-causality, there was evidence of a bidirectional causality between: population to energy consumption, industrialization to CO2, industrialization to trade opennes, population to trade openess, GDPPC to trade opennes, and population to industrialization. Moreover, there was evidence of a unidirectional causality running from CO2 to energy consumption, trade opennes to CO2, population to CO2, GDPPC to industrialization, and GDPPC to population. Evidence from the Cholesky’s method of variance decomposition shows that 13.13% of future fluctuations in energy consumption are due to shocks in Carbon dioxide emissions, 33.63% of future fluctuations in carbon dioxide emissions are due to shocks in energy consumption, 32.88% of future fluctuations in trade openness are due to shocks in industrialization, 40.63% of future fluctuations in industrialization are due to shocks in energy consumption, 41.23% of future fluctuations in population are due to shocks in energy consumption, and 27.99% of future fluctuations in GDPPC are due to shocks in trade openness. Several policy implications can be drawn based on empirical findings concerning energy consumption in Somalia. To begin with, diversification of Somalia’s economic productivity through enhanced technological advancement; innovation; value added to raw materials, goods, and services will broaden the financial base leading to high levels of per capita GDP. Furthermore, diversification of Somalia’s energy sector throughout the incorporation of renewable energy technologies into the energy portfolio will reduce reliance on volatile energy sources (fuel, oil, and gas) and the likelihood of interruptions in energy supply due to commodity price volatility, thereby contributing to climate change mitigation efforts. REFERENCES African Development Bank. (2015), Somalia Energy Sector Needs Assessment and Investment Programme. Abidjan: African Development Bank. Available from: https://www.afdb.org Asumadu-Sarkodie, S., Owusu, P.A. (2016a), The potential and economic viability of solar photovoltaic power in Ghana. Energy Sources, Part A: Recovery, Utilization, and Environmental Effects, 38(5), 709-716. Asumadu-Sarkodie, S., Owusu, P.A. (2016b), Energy use, carbon dioxide emissions, GDP, industrialization, financial development, and population, a causal nexus in Sri Lanka: With a subsequent prediction of energy use using neural network. Energy Sources. Part B Economics, Planning and Policy 11(9):889-899. Asumadu-Sarkodie, S., Owusu, P.A. (2017), The causal nexus between energy use, carbon dioxide emissions, and macroeconomic variables in Ghana. Energy Sources, Part B: Economics, Planning and Policy, 12(6), 533–546. Banerjee, A., Dolado, J.J., Mestre, R. (1998a), Error-correction mechanism tests for cointegration in a single-equation framework. Journal of Time Series Analysis, 19(3), 267-283. Banerjee, A., Dolado, J.J., Mestre, R. (1998b), Error-correction mechanism tests for cointegration in a single-equation framework. Journal of Time Series Analysis, 19(3), 267-283. Boswijk, H.P. (1994), Testing for an unstable root in conditional and structural error correction models. Journal of Econometrics, 63(1), 37-60. Cerdeira Bento, J.P., Moutinho, V. (2016), CO2 emissions, non-renewable and renewable electricity production, economic growth, and international trade in Italy. Renewable and Sustainable Energy Reviews, 55, 142-155. Chandia, K.E., Gul, I., Aziz, S., Sarwar, B., Zulfiqar, S. (2018), An analysis of the association among carbon dioxide emissions, energy consumption and economic performance: An econometric model. Carbon Management, 9(3), 227-241. Chen, Y., Wang, Z., Zhong, Z. (2019), CO2 emissions, economic growth, renewable and non-renewable energy production and foreign trade Warsame, et al.: The Causal Relation between Energy Consumption, Carbon Dioxide Emissions, and Macroeconomic Variables in Somalia International Journal of Energy Economics and Policy | Vol 13 • Issue 3 • 2023110 in China. Renewable Energy, 131, 208-216. Dong, K., Hochman, G., Zhang, Y., Sun, R., Li, H., Liao, H. (2018), CO2 emissions, economic and population growth, and renewable energy: Empirical evidence across regions. Energy Economics, 75, 180-192. Edenhofer, O., Madruga, R.P., Sokona, Y., Seyboth, K., Matschoss, P., Kadner, S., Zwickel, T., Eickemeier, P., Hansen, G., Schlömer, S., von Stechow, C. (2011), Renewable energy sources and climate change mitigation: Special report of the intergovernmental panel on climate change. In: Renewable Energy Sources and Climate Change Mitigation: Special Report of the Intergovernmental Panel on Climate Change. United Kingdom: Cambridge University Press. p1-1075. Kunvitaya, A., Dhakal, S. (2017), Household energy requirements in two medium-sized Thai cities with different population densities. Environment and Urbanization, 29(1), 267-282. Lin, B., Omoju, O.E., Okonkwo, J.U. (2015), Impact of industrialisation on CO2 emissions in Nigeria. Renewable and Sustainable Energy Reviews, 52, 1228-1239. Lipton, M., Ravallion, M. (1993), Poverty and policy. Handbook of development economics, 3, 2551-2657. Lovins, A.B. (1990), The effect of urbanization and industrialization on energy use in emerging economies: Implications for sustainable development. Population and Development Review, 16, 95. Mohiuddin, O., Asumadu-Sarkodie, S., Obaidullah, M. (2016), The relationship between carbon dioxide emissions, energy consumption, and GDP: A recent evidence from Pakistan. Cogenteng Engineering, 3, 1. Ohlan, R. (2015), The impact of population density, energy consumption, economic growth and trade openness on CO2 emissions in India. Natural Hazards, 79(2), 1409-1428. Otsuka, A. (2018), Population agglomeration and residential energy consumption: Evidence from Japan. Sustainability, 10(2), 469. Rafindadi, A.A., Ozturk, I. (2015), Natural gas consumption and economic growth nexus: Is the 10th Malaysian plan attainable within the limits of its resource? Renewable and Sustainable Energy Reviews, 49, 1221-1232. Salahuddin, M., Gow, J., Ozturk, I. (2015), Is the long-run relationship between economic growth, electricity consumption, carbon dioxide emissions and financial development in Gulf Cooperation Council Countries robust? Renewable and Sustainable Energy Reviews, 51, 317-326. Sarkodie, S.A., Adams, S. (2018), Renewable energy, nuclear energy, and environmental pollution: Accounting for political institutional quality in South Africa. Science of the Total Environment, 643, 1590-1601. Shahbaz, M., Khraief, N., Jemaa, M.M.B. (2015), On the causal nexus of road transport CO2 emissions and macroeconomic variables in Tunisia: Evidence from combined cointegration tests. Renewable and Sustainable Energy Reviews, 51, 89-100. Shahbaz, M., Lean, H.H. (2011), Does financial development increase energy consumption? The role of industrialization and urbanization in Tunisia. Energy Policy, 40, 473-479. Shpak, N., Ohinok, S., Kulyniak, I., Sroka, W., Fedun, Y., Ginevičius, R., Cygler, J. (2022), CO2 Emissions and macroeconomic indicators: Analysis of the most polluted regions in the world. Energies, 15(8), 15082928. Somalia Energy Statistics-Worldometer. (n.d.), Available from: https:// www.worldometers.info/energy/somalia-energy [Last accessed on 2023 Feb 15]. Somalia Power Africa U.S. Agency for International Development. (n.d.), Available from: https://www.usaid.gov/powerafrica/somalia [Last accessed on 2023 Feb 15]. Warsame, A.A., Sarkodie, S.A. (2021), Asymmetric impact of energy utilization and economic development on environmental degradation in somalia. Environmental Science and Pollution Research,  29(16), 23361-23373. Warsame, A.A., Sheik-Ali, I.A., Mohamed, J., Sarkodie, S.A. (2022), Renewables and institutional quality mitigate environmental degradation in Somalia. Renewable Energy, 194, 1184-1191. Warsame, Z.A. (2023), The significance of FDI inflow and renewable energy consumption in mitigating environmental degradation in Somalia. International Journal of Energy Economics and Policy, 13(1), 443-453. World Bank. (n.d.), CO2 emissions (kt)-Somalia Data. Available from: https://www.data.worldbank.org/indicator/EN.ATM.CO2E. KT?locations=SO [Last accessed on 2023 Mar 04].