Impact of Electricity Loss on Gross Domestic Product in Nigeria American Interdisciplinary Journal of Business and Economics ISSN: 2837-1909| Impact Factor : 4.6 Volume. 9, Number 2; April-June, 2022; Published By: Scientific and Academic Development Institute (SADI) 8933 Willis Ave Los Angeles, California https://sadipub.com/Journals/index.php/aijbe 1 American Interdisciplinary Journal of Business and Economics | https://sadipub.com/Journals/index.php/aijbe IMPROVING TRANSMISSION AND DISTRIBUTION INFRASTRUCTURE TO REDUCE ELECTRICITY LOSS AND BOOST GDP IN NIGERIA Akindele Samson Department of Economics, Trinity University, Lagos, Nigeria Abstract: This study examines the impact of electricity loss on the economic growth of Nigeria from the period 1981 to 2020. The study employs a Vector Autoregressive (VAR) model to estimate the effects of electricity loss, electricity demand, and electricity supplied from hydropower on the Gross Domestic Product (GDP) of the country. The results of the study reveal that electricity loss has a negative impact on the GDP, while electricity demand has a positive impact on GDP, and electricity supplied from hydropower negatively affects GDP. However, the study finds that there is no causal relationship between electricity loss, electricity demand, and electricity supply, and GDP in Nigeria. The study recommends a deliberate policy to stimulate investment in the transmission and distribution infrastructure to reduce electricity losses in the country. The study highlights the different forms of electricity losses, which can be technical or non-technical, and categorizes them into transmission and distribution losses. The study suggests that decreasing electricity losses leads to an increase in GDP through an increase in distribution company revenue and an increase in the quantity of electricity reaching the end-users. Keywords: electricity loss, economic growth, Gross Domestic Product (GDP), Vector Autoregressive (VAR) model, transmission and distribution losses, Nigeria. Introduction Electricity is a vital input for economic growth and development in all sectors of the economy. However, it is a well-known fact that many countries, especially developing countries, experience significant losses in the electricity sector due to various reasons such as outdated infrastructure, vandalism, and theft. These losses not only impact the quality and availability of electricity to households and businesses but also impede the economic growth of these countries. Nigeria, being the most populous country in Africa, has experienced frequent electricity outages that have had severe consequences on the economy. This study aims to investigate the impact of electricity loss on the Gross Domestic Product (GDP) of Nigeria from 1981 to 2020, using a Vector Autoregressive model. The study considers electricity loss, electricity demand, and electricity supplied from hydropower as the primary variables affecting GDP. The study provides policy recommendations to reduce electricity losses and increase economic growth in Nigeria. The remainder of the study is divided into five sections. The next section presents a review of the empirical literature on electricity loss and economic growth. The third section outlines the theoretical framework underpinning this study, while Section four presents the methodology and model specification. Section five provides sources and data for the study, and the final section presents and discusses the results alongside the policy recommendations for reducing electricity losses and increasing economic growth in Nigeria. There are different degrees of electricity loss along the transmission lines. Chint (2020) showed that as the electricity is being generated and stepped up by the transformer, about 1 to 2 percent of generated energy is lost. Another 2 to 4 percent is again lost in the transmission lines, while another 1 to 2 percent is lost during step down from the high tension wires. More importantly, about 4 to 6 percent is lost in the distribution process Akindele Samson (2022) 2 American Interdisciplinary Journal of Business and Economics | https://sadipub.com/Journals/index.php/aijbe to the end users. On the whole, about 8 to 15 percent of electricity is lost between the power plants and the end users. This overall percentage loss implies a heavy dampening effect on Gross Domestic Product through firms’ productivity. Dakpogan and Smit (2018) estimated the effect on GDP from 0.5 to 1.2 percent in some Sub-Saharan African countries. A comparison of electricity loss in Nigeria with some other countries can be observed in figure one. The country with the highest volatility of losses is Nigeria followed by Ghana. The highest for electricity loss in Britain was 9.8 percent in 1994 while the highest loss in Nigeria was 49.27 percent in 1981. The lowest loss that Nigeria has ever had was 4.5 in 2009. This amount of electricity loss in Nigeria is quite worrisome and could be a factor responsible for the incessant power outages in the country and this has grave consequences on productivity in the country and attendant effect on GDP growth. Figure 1: Electricity Loss in some selected countries Source: Author generated using Excel 2010 Losses in electricity transmission and distribution in the Latin American and the Caribbean countries was estimated at between 0.19 and 0.3 of GDP for all the countries and the losses in terms of monetary value ranges between $11 to $17 billion (Jiménez, Serebrisky and Mercado, 2014). On a country by country analysis, electricity loss in Brazil cost about 0.26 of GDP, while that of Mexico is between 0.12 and 0.3 (Jiménez, et al, 2014). These authors estimated technical losses in Uruguay at $80 million as at 2010. The implication of these is that as the monetary losses increase, it finds its expression in reduction of productivity and consequently on economic growth. A reduction in electric power losses leads to increase in GDP through increase in the revenue of the distribution companies, as a result of increase in the quantity of electricity that reaches the end users. This increase in electricity supply will increase its consumption and a reduction in the use of alternative sources of electricity, like generators, which in turn reduces the cost of production and raises profits of firms and expands investment potentials. This work is particularly important because, to the best of the author’s knowledge, most of the works done on electricity loss in Nigeria have only considered the effect of the loss on power outages in the country using descriptive analysis without a sound methodological approach. None of the works have investigated the impact of electricity loss on the Gross Domestic Product of the country. The remaining part of this work will be divided into six sections. The following section will handle the recent empirical literature, while the following section three will present the theoretical framework underpinning this study. Section four will show the methodology and the model specification while section five will show the sources of data for the work and section six will present the result and then analyze. Section seven will then summarize and give policy recommendations. Literature Review Das and McFarlane (2021) investigated the effect of remittances on electricity consumption and electricity losses in Jamaica between 1976 and 2014. The work employed Vector Error Correction model and Granger causality to analyze the work. The findings revealed that cointegration existed between remittances and electricity consumption and losses. It was also discovered that there was a bi-directional causality among the variables; however, the relationship between electricity losses and remittances was negative while that between energy consumption and remittances was positive. It was therefore recommended that social policies be designed to make use of remittances to reduce electricity theft in the country. Adams, Atsu, Klobodu and Lamptey (2020) employed Autoregressive Distributed Lag model to evaluate the effect of transmission and distribution losses on the growth of South African economy using time series data Akindele Samson (2022) 3 American Interdisciplinary Journal of Business and Economics | https://sadipub.com/Journals/index.php/aijbe between 1971 and 2014. Their findings showed that there existed a long run relationship among the variables by using Foreign Direct Investment as a control variable. By the time the robustness check was carried out it was found that 1 percent change in electricity loss in the country results in a decline in economic growth from about 3.8 percent to 2.2 percent. It was therefore recommended that policy makers should invest more in energy production and infrastructure to help reduce electricity loss due to technical factors. George-Anokwuru and Ekpenyong (2020) did an empirical study on relationship between electricity and economic growth in Nigeria between 1971 and 2018 and using consumption of electricity, generation along with transmission and distribution losses. The work employed Autoregressive Distributed Lag model, the findings revealed that electricity generation and distribution losses have negative relationship with economic growth and electricity generation has a negative relationship with economic growth. It was therefore suggested that production and transmission infrastructure expenditure be encouraged to boost energy consumption and reduce the losses. Dakpogan and Smit (2018) looked at the effect of electricity loss on Gross Domestic Product (GDP) in Benin Republic between 1980 and 2014 by employing the methodology of Autoregressive Distributed Lag model. The findings of the study showed that a 1 percent loss in electricity in the country would cost about 0.16 percent decline in GDP. It was therefore proposed that a financing mechanism on electricity infrastructure to reduce the losses from electricity in order to increase GDP be designed and implemented. Adams, Klobodu and Lamptey (2017) carried out an examination of how electricity loss from transmission to distribution affects GDP growth in Ghana between 1971 and 2012. The work employed Autoregressive Distributed Lag model using Bounds test for cointegration and their findings revealed that there was a long run relationship between electricity loss and economic growth in the country. It was further discovered that electricity loss did not show any significant effect on economic growth but when the urban population growth was factored in, it was discovered that electricity loss became significant in affecting economic growth in Ghana. Costa-Campi, Daví-Arderius and Trujillo-Baute (2016) carried out an analysis of the impact of consumption and generation on electricity losses in Spain. The study used a quantitative analysis of the marginal effect of losses in Mega Watts from an additional Mega Watts produced or consumed. It was found that there is a huge amount of saving that will be made when losses can be reduced by using smart meters. It was found out from the result that a 1 percent reduction in loss of electricity will lead to about 1.25 percent in saving. It was also discovered that in increase in solar and wind capacity would reduce energy losses. It was therefore recommended that a stronger coordination between the transmission and distribution system operators be encouraged. Theoretical Framework The work of Samuelson and Nordhaus (1989) on growth will be the framework on which this study will rest. The mainstream theory of economic growth claims that production plays the most important role of determining the growth of an economy, and every production process requires energy to be able to transform any matter into finished product. The theory classifies capital, labour and land into primary factors that must be obtained at the beginning of any production and they are not used up although they can be degraded or improved upon. But the theory asserts that energy resources or commodities like oil, gas fuels, electricity and coal are classified as intermediate inputs and can be completely used up in the production process. This theory helps us to explain that production can be influenced by the availability of electricity. The more the productive agents have access to electricity the more productivity can be enhanced. Increase in energy loss will imply shortage of electricity supply and consequently the negative implication on productivity and growth. Methodology and Model Specification This study employed an ex post facto research design by trying to interrogate the relationship between electricity loss and economic growth in Nigeria. An econometric method of Vector Autoregressive (VAR) model will be employed. The reason is that this method of estimation is to be used when all the variables are stationary at first difference or integrated at order one, I(1). The data were subjected to unit root test using Augmented Dickey Fuller test method and it was found that all variables were integrated at I(1). The first step after confirming the order of integration is to test for the long run relationship among the variables and this was done using Johansen cointegration test and it was found that there was no cointegration, and so the work has to employ Vector Autoregressive model. Akindele Samson (2022) 4 American Interdisciplinary Journal of Business and Economics | https://sadipub.com/Journals/index.php/aijbe The model for this work is hinged on the growth model proposed by Samuelson and Nordhaus (1989) where the Cobb-Douglas production function depends on labour and capital and energy as specified below: Output = f (Labour, Capital, Energy) … (1) Qt = wLa CB E1-a-B … (2) Where is the output, is the total factor productivity, L is the labour input, C is the capital input and E is the energy input, , and are the share of the factor inputs in income. Adapting model (2) and incorporating the electricity components that is the focus of this study, we then have: GDP=f(ELELOS,ELEDD,ELEHP) … (3) Expressing (3) econometrically, we have: RGDPt = a + B1ELELOSt + B2ELEDDt + B3ELEHPt + ut … (4) Where RGDP is the real Gross Domestic Product, ELELOS is the electricity loss both at the transmission and distribution lines, ELEDD is electricity demand in total and ELEHP is electricity supplied from hydropower source, is the error term. 4.1 Specification of Vector Autoregressive Model The basic form of a VAR model consists of a set of M variables that are endogenous: gt = g1t , ..,gmt, For m = 1, …M. When the lags p of the endogenous variables is included, we have VAR(p) expresses as: gt = B1gt-1 + … + Bpgt-p + EFt + Et … (5) Where are (M x M) coefficient matrices for i =1, …., p, is an M-dimensional white noise with time invariant positive definite covariant matrix , where E[Et] = 0 whose value is positive and definite The matrix E is the matrix of coefficients of possible deterministic regressors having dimension (M x N), while F is a column vector containing deterministic regressors like constant, trend, dummy and seasonal variables. Equation (5) can be expressed VAR with the variables of interest in this work as: 𝑅�𝐺�𝐷�𝑃�𝑡��=�∑𝒌�𝒊�=𝟏��𝛽�𝑖��𝑅�𝐺�𝐷�𝑃�𝑡�−1�+ ∑𝒌�𝒋�=𝟏��𝛽�𝑗��𝐸�𝐿�𝐸�𝐿�𝑂�𝑆�𝑡�−1 + ∑𝒌�𝒏�=𝟏��𝛽�𝑛��𝐸�𝐿�𝐸�𝐷�𝐷�𝑡�−1 + ∑𝒌�𝒎�=𝟏��𝛽�𝑚���𝐸�𝐿�𝐸�𝐻�𝑃�𝑡�−1�+� u1t … (6) … (7) 𝐸�𝐿�𝐸�𝐷�𝐷�𝑡��=�∑𝒌�𝒊�=𝟏��𝛽�𝑖��𝑅�𝐺�𝐷�𝑃�𝑡�−1�+ ∑𝒋�𝒌�=𝟏��𝛽�𝑗��𝐸�𝐿�𝐸�𝐿�𝑂�𝑆�𝑡�−1 + ∑𝒌�𝒏�=𝟏��𝛽�𝑛��𝐸�𝐿�𝐸�𝐷�𝐷�𝑡�−1 + ∑𝒌�𝒎�=𝟏��𝛽�𝑚���𝐸�𝐿�𝐸�𝐻�𝑃�𝑡�−1 + u3t … (8) 𝐸�𝐿�𝐸�𝐻�𝑃�𝑡��=�∑𝒌�𝒊�=𝟏��𝛽�𝑖��𝑅�𝐺�𝐷�𝑃�𝑡�−1�+ ∑𝒋�𝒌�=𝟏��𝛽�𝑗��𝐸�𝐿�𝐸�𝐿�𝑂�𝑆�𝑡�−1 + ∑𝒌�𝒏�=𝟏��𝛽�𝑛��𝐸�𝐿�𝐸�𝐷�𝐷�𝑡�−1 + ∑𝒌�𝒎�=𝟏��𝛽�𝑚���𝐸�𝐿�𝐸�𝐻�𝑃�𝑡�−1 + u4t … (9) where the variables remained as defined earlier. Data and Sources Annual time series data will be used for this study and it will cover the period between 1981 and 2014. Real Gross Domestic Product data is obtained from the National Bureau of Statistics of the country while data for Electricity Loss, Electricity demand, and Electricity from hydropower are all sourced from the World Bank Development Indicator. Electricity loss, electricity demand and supply are expressed as percentage of output. Table 1: Variables, Definition and Source No Variable Definition Source 1 RGDP Real Gross Domestic product National Bureau of Statistics 2 ELELOS Electricity Loss WDI 3 ELEDD Elecricity Demand WDI 4 ELEHP Elecridity supply from Hydropower WDI Note: WDI is World Development Indicator Data base of World Bank Akindele Samson (2022) 5 American Interdisciplinary Journal of Business and Economics | https://sadipub.com/Journals/index.php/aijbe Results and Discussion 6.1 Unit Root Test The variables were tested for the presence of unit root, because time series are assumed to be stationary. Table 2: Stationarity Test (Augmented Dickey Fuller) At Level At First Difference Variable ADF Stat 5 % Level Prob. Value ADF Stat 5 % Level Prob. Value Order of Integration RGDP 0.650415 -2.92245 0.9898 -5.95836 -2.92378 0.0000 I(1) ELELOS -1.45922 -2.92517 0.5453 -6.54566 -2.92517 0.0000 I(1) ELEDD -1.48747 -2.92245 0.5316 -9.39097 -2.92378 0.0000 I(1) ELEHP -2.70943 -2.92245 0.0797 -8.98493 -2.92378 0.0000 I(1) Source: Author generated using Eviews 10 The result of the test is in table 1. It showed that all the variables are stationary at first difference with all the probability values less than 5 percent. 6.2 Optimum Lag Selection The lag selection criteria were used and all the criteria selected lag length one as indicated in table 3. Table 3: Optimum lag length selection Lag LogL LR FPE AIC SC HQ 0 -1080.21 NA 4.89e+14 45.17536 45.33129 45.23428 1 -908.892 306.9428* 7.59e+11* 38.70382* 39.48348* 38.99845* 2 -895.812 21.25386 8.70e+11 38.82551 40.22891 39.35586 *Lag length selected. Source: Author generated using Eviews 10 6.3 Cointegration test From the result of the unit root test results, the test for the existence of long run relationship among the variables was carried out using Johansen cointegration test method as shown in table 3. Table 4: Cointegration (Johansen) Trace Test Max-Eigen Test Null: No of Coint. 5 % Prob. Trace Statistic Level Value 5 % Max-Eigen Level Prob. Statistic Value None 32.62383 47.85613 0.5777 14.44681 27.58434 0.7906 At most 1 18.17702 29.79707 0.5530 10.56215 21.13162 0.6907 At most 2 7.614869 15.49471 0.5074 7.600756 14.26460 0.4207 Source: Author generated using Eviews 10 The condition for accepting the null hypothesis is that if the probability value of the first null hypothesis is greater than 5 percent critical value, we accept the null hypothesis that there is no cointegration. The probability is 79 percent, so we conclude that there is no cointegration among the variables and we can only run the VAR at level. 6.4 VAR Impulse Response Function The VAR models (6) to (9) were estimated and the impulse response functions were generated as shown in figure 2 to 4. Figure 2: Response of RGDP to electricity loss Akindele Samson (2022) 6 American Interdisciplinary Journal of Business and Economics | https://sadipub.com/Journals/index.php/aijbe Source: Author generated using Eviews 10 The response of GDP to a shock in electricity loss as shown in figure 2 assumes a negative dimension from period one to period 10 of the horizon. This implies that GDP and electricity loss are negatively related and significant. It also means that as more electricity is lost through the technical and non-technical channels, it is manifesting a dampening effect on the nation’s productivity. This finding is in support of most of the works from other countires, such as Dakpogan and Smit (2018) and George-Anokwuru and Ekpenyong (2020). Figure 3 on the other hand showed the response of GDP to a shock in electricity demand. It showed that GDP is positively responding to electricity demand. This implied that if more electricity is available for consumption it has a positively strong impact on GDP. It is obvious that the more the available electricity for consumption, the more positive ripple effect it will have on GDP of the country. The response of GDP to a shock in electricity supply from hydropower is shown in figure 4 and it revealed that GDP is negatively related to electricity supply from hydropower. It is expected however, since electricity loss is huge, it means the available electricity supplied will not be able to influence GDP significantly. Figure 3: Response of RGDP to electricity demand Source: Author generated using Eviews 10 Figure 4: Response of RGDP to electricity supplied from hydropower Source: Author generated using Eviews 10 6.5 Residual Diagnostics The VAR model was checked if the assumptions of Ordinary Least Squares are satisfied. Table 4 showed that the model is free from serial correlation among the residuals and that is makes the model suitable for policy Akindele Samson (2022) 7 American Interdisciplinary Journal of Business and Economics | https://sadipub.com/Journals/index.php/aijbe prescription and forecast. But the errors in the model are not free from heteroscedasticity and they are not normally distributed. Table 5: Residual Diagnostic tests VAR Residual Serial Correlation LM Tests Lag LRE stat df Prob. Rao F-stat df Prob. value 1 21.23046 16 0.1698 1.366627 (16, 13.7) 0.1711 VAR Residual Heteroskedasticity Tests (Levels and Squares) Joint test: Chi-sq df Prob. 139.5899 80 0.0000 VAR Residual Normality Tests Component Jarque-Bera df Prob. RGDP 701.0944 2 0.0000 ELELOS 3.775582 2 0.1514 ELEDD 3.806633 2 0.1491 ELEHP 33.81916 2 0.0000 Joint 742.4958 8 0.0000 Source: Author generated using Eviews 10 6.6 Granger Causality test It was shown in table 5 that none of the independent variables could cause GDP with all their probability values more than 5 percent significant level. This implies that even though electricity loss is negatively impacting GDP, there is no causality running from electricity loss to GDP. Similarly, electricity demand and supplied from hydropower could not Granger cause GDP. Table 6: VAR Granger Causality test Dependent variable: RGDP Excluded Chi-sq df Prob. ELELOS 0.238600 1 0.6252 ELEDD 1.609738 1 0.2045 ELEHP 0.335856 1 0.5622 All 3.975535 3 0.2641 Source: Author generated using Eviews 10 Conclusion and Policy Implications This study investigated the impact of electricity loss on Real Gross Domestic Product in Nigeria along with electricity demand and supplied from hydroelectric power. It was found that there was no long run relationship among the variables, Vector Autoregressive impulse response analysis was employed to check the how the independent variables stimulate the response of RGDP and it was revealed that GDP responds negatively to electricity loss significantly through the period under consideration. Although the period covered was limited due to availability of data on electricity loss, the situation described in this work has not changed, but rather grown worse. Similarly, electricity demand impacted GDP positively and electricity supplied from hydropower impacted GDP negatively. This negative relationship between electricity demand and GDP shows that if only more electricity could be supplied, it would boost GDP. It was also shown that electricity supply showed a negative impact on GDP simply due to the huge loss of electricity through transmission and distribution channels. The major recommendation therefore is that there should be a sound energy policy that would be designed to encourage investment in the transmission and distribution lines to upgrade the infrastructure with the purpose of reducing electricity loss. It is certain that if the losses can be reduced, it will increase the quantity being Akindele Samson (2022) 8 American Interdisciplinary Journal of Business and Economics | https://sadipub.com/Journals/index.php/aijbe supplied and consequently increase the quantity demanded and eventually translate to higher productivity and GDP. References Adams, S., Atsu, F., Klobodu, E. M. and Lamptey, R. (2020). Electricity transmission, distribution losses and economic growth in South Africa. Heliyon 6, 1-9. Adams, S., Klobodu, E. K. M. and Lamptey, R. O. (2017). Electric Power Transmission, Distribution Losses, and Economic Growth in Ghana. Social, Health, and Environmental Infrastructures for Economic Growth, 18. Antmann, P. (2009). Reducing technical and non-technical losses in the power sector. Background Paper for the World Bank Group Energy Sector Strategy. Available at: https://siteresources.worldbank.org/EXTESC/Resources/Background_paper_Reducing_losses_in_the _p ower_ Bandim, C. Alves Jr., J. Pinto Jr., A. Souza, F. Loureiro, M. Magalhaes, C. and Galvez-Durand, F. (2003). “Identification of energy theft and tampered meters using a central observer meter: a mathematical approach,” in IEEE PES Transmission and Distribution Conference and Exposition, Dallas, 2003, 163- 168. Chint (2020). https://chintglobal.com/blog/how-much-power-loss-in-transmission-lines/ Costa-Campi, M. T., Daví-Arderius, D. and Trujillo-Baute, E. (2016). The Economic Impact of Electricity Losses. Documents de Treball de l’IEB 2016/4. Dakpogan, A. and Smit, E. (2018). The effect of electricity losses on GDP in Benin. Munich Personal RePEc Archive (MPRA), Paper No. 89545 Das, A. and McFarlane, A. (2021), “Remittances, electricity consumption and electric power losses in Jamaica”, Journal of Economic Studies, https://doi.org/10.1108/JES-09-2020-0466 Eberhard, A,, Foster, V., Briceño-Garmendia, C., Ouedraogo, F., Camos, D. and Shkaratan, M. (2008). Underpowered: the state of the power sector in Sub-Saharan Africa. AICD Background Paper 6, World Bank. George-Anokwuru, C. C. and Ekpenyong, B. I. (2020). Electricity and Economic Growth in Nigeria. Journal of Economics, Business and Market Research, 1(1), 1-7. International Monetary Fund (IMF). (2015). Structural reforms and macroeconomic performance: initial considerations for the Fund. IMF policy paper, staff report. Washington, D.C. Jiménez,R., Serebrisky, T. and Mercado, J. (2014). Sizing Electricity Losses in Transmission and Distribution Systems in Latin America and the Caribbean, Inter-American Development Bank Felipe Herrera Library. 1-42. Jones, C, I. (2011). Intermediate goods and weak links in the theory of economic development. America Economic Journal of Macroeconomics. 3, 1–28. Nagi, J., Yap, K., Nagi, F., Tiong, S. Koh, S. and Ahmed, S. (2010). “NTL detection of electricity theft and abnormalities,” in Proceedings of 2010 IEEE Student Conference on Research and Development (SCOReD 2010), Putrajaya, Malaysia, 2010, 202-206 Akindele Samson (2022) 9 American Interdisciplinary Journal of Business and Economics | https://sadipub.com/Journals/index.php/aijbe Nigerian Electricity Regulation Commission (NERC), (2019). “Industry statistics” [Online]. Available: https://nercng.org/index.php/library/industrystatistics/distribution/119-atc-c-losses/464- phdisco#data. [Accessed 20 March, 2019). Ogujor, E. and Otasowie, P. (2010). “The impact of the pre-paid meter on revenue generation in Nigeria,” Pacific Journal of Science and Technology, 11(1), 138-142. Payne, J. E. (2010). A survey of the electricity consumption-growth literature. Applied Energy, 87(3), 723– 731. http://doi.org/10.1016/j.apenergy.2009.06.034 Samuelson, P. A. and Nordhaus, W. D. (1989). Economics. McGraw-Hill, 1989 World Development Indicators. (2017). https://data.worldbank.org/products/wdi http://doi.org/10.1016/j.apenergy.2009.06.034 http://doi.org/10.1016/j.apenergy.2009.06.034