The Determinants of Energy and Electricity Consumption in Developed and Developing Countries: International Evidence

Abstract

Aim and background—As research on the energy and electricity consumption determinants yields mixed results and a multifactorial model has not yet been developed, our study aims to investigate the growth dynamics of the factors that affect energy consumption in developed and developing countries. Motivation—The current global energy crisis has led us to a thorough investigation of the determinants that are affecting it. Hypothesis—It is hypothesized that a set of macro-financial, macro-environmental, and institutional variables are factors that causally affect energy and electricity consumption in a holistic model. Methods—This research uses the data from 109 countries within a multivariate panel framework taken during 2010–2018, through the error correction, dynamic cointegration econometric methodologies, and causality tests. Results—The results indicate a coherent model with high interpretive power (80%) and that the main determinants of energy consumption in developing countries are economic growth, investment, and winter temperature, whereas, in developed countries, the determinants are trade openness, corruption, and innovation. Conclusion—Because energy consumption and economic growth share a bilateral relationship, the conservation of energy policy measures must be implemented according to the income category in which the country is classified.

1. Introduction

Increased demand for energy, limited natural resources, a strong commitment to reducing climate pollution, technological development, and various political factors seem to be some of the causes of the recent crisis in energy prices worldwide [1,2]. With the rapid economic growth and technological progress, the demand for energy in developed and developing countries has increased significantly. Although energy consumption can lead to economic growth, it is the primary cause of environmental degradation and all the problems related to it [3]. Great efforts are currently being made to efficiently use energy sources in many developed countries and a shift to renewable energy sources is also seen. However, in developing countries, an insufficient energy supply would be detrimental to economic growth. Some characteristics of such economies are their overreliance on the external economies for the energy supply, the usage of energy-inefficient devices, and frequent unannounced power cuts [4].

Although a considerable amount of literature on energy consumption, economic growth, and the environment has been developed over the years, little consensus has been reached yet about the factors that affect energy consumption. Therefore, recent studies have highlighted the need for new innovative research that ought to initially focus on creating a holistic model correlating energy consumption with the key factors that affect it [5,6,7]. Creating a comprehensible and acceptable model could lead to policy decisions that can desirably modify the consequences. To date, the efforts of the previous research are fragmentary in nature as they have taken into account only a few individual factors or case studies solely based on households [8], regions [9], a single country, or a limited number of countries [10,11]. Several studies have dealt with the interactions between neighboring countries and spatial concentration on energy consumption based on their economic sectors [12]. This gap in the literature will be addressed in the present study as this study utilizes a multifactorial model with several independent variables with a total of 109 developing and developed economies under consideration.

Because of the increase in energy consumption in the face of limited supply, the researchers review empirical work to determine the main variables responsible for energy demand to contribute to the existing literature on the determinants of energy consumption, especially because the current research is not exhaustive and exhibits inconsistencies with their results [13]. Hence, more reviews on the recent empirical studies are essential to inform policymakers, students, and researchers in energy and sustainable development [10,14]. This research aims to contribute to the body of knowledge in the literature on energy consumption and its determinants by identifying the various factors that affect the energy demand at the macroeconomic level. It also examines the elasticities of variables as its knowledge is relevant for policymaking in energy management.

The factors that lead to the current energy crisis in energy demand and supply were studied. The first key factor is the increase in demand due to a country’s economic and financial development—for which the key variables are the gross domestic product per capita (GDPpc) and financial development [15,16,17,18,19,20]. At the same time, the excessive use of existing natural resources (oil, gas, and coal) that exists in limited capacities exacerbates the problem. The usage of renewable energy sources can reduce the dependence on fossil fuels and help reduce greenhouse gas emissions, yet they are still largely unused in most countries [21,22]. Another cause for the crisis was the ever-increasing global population and its demand for fuel and products. None of the types of food or products used are produced or transported without the significant depletion of energy resources. The proxy of this variable is population growth [18,23]. In addition, an essential factor is the ageing of power generation equipment and infrastructure. Most energy companies and households continue to use outdated equipment that reduces energy production. Not only would investments in the whole economy be then required to upgrade the infrastructure of production as well as the use of energy, but also innovative practices in the entire range of production, distribution, and use of resources would become critical [24,25].

One of the main goals of the current research is to investigate the causality between the GDPpc and energy consumption through different dynamic econometric processes. Results are then extracted in order to confirm them with varying samples of country and different variables through causality tests prescribed by Dumitrescu and Hurlin [26]. The purpose is to clarify the direction and the intensity of the relationship by shedding light on this disputed correlation. At the same time, to improve the current literature, we added to the single model a set of macro-financial variables (GDPpc, investment, imports-exports, and financial development), macro-environmental variables (winter temperatures, energy losses, and renewable energy), institutional variables (corruption and political stability), and other variables (population growth and innovation) to investigate their combined effect on energy and electricity consumption. Because these variables are cross-sectional and dependent on each other, appropriate econometric methodologies have been used to handle this problem [27,28,29]. Corresponding attempts with multivariate models have been made on a limited scale with dependent variables such as CO₂ emissions and the ecological footprint. The ultimate goal is to identify the correlations that lead to increased energy consumption—which is one of the two main forces (or demands) behind the current energy crisis.

In many countries, particularly the developing countries, there is a significant delay in the launch of new power plants that can bridge the gap between energy supply and demand or in the issuance of licenses for renewable energy sources. A key reason for this is bureaucracy and corruption inherent within such structures, which will be the variable used in the present study [10,30,31,32]. In addition, people often overlook the importance of saving energy. At the same time, there are many leaks in the production and distribution of energy to users. Thus, the energy loss variable will be used [33]. Moreover, extreme weather conditions in winter or summer significantly affect the demand and supply of energy [5,34]. Finally, over-taxation, political instability, and wars can affect the energy balance in individual countries as well as around the world [17,35,36,37].

The primary purpose of this research was to determine a holistic energy consumption model, which has been neglected in the existing literature. Many econometric models (such as panel-corrected standard error (PCSE), autoregressive factor (AR1), fully modified ordinary least square (FM-OLS) and dynamic-ordinary least square (D-OLS)) were incorporated for this purpose. Empirical evidence showed that a 1% increase in the GDPpc increases energy consumption by 0.6 to 0.65%. In addition, in our research, the differences in the sample of developing and developed countries concerning the impact of GDPpc, corruption, and financial development on the energy consumption model have been identified and analyzed. One of the main novelties of the current study is that it has been able to identify the different policy decisions that need to be taken to address the problem of high energy consumption (and consequently the energy crisis and environmental degradation) in developing and developed countries. Developing countries need intervention in tackling corruption, liberalizing trade, and investing in energy transmission systems, and developed countries could benefit from the interventions in total renewable energy investments and the quality of the workforce.

We also found that an increase in corruption and private sector financing causes an increase in energy consumption. At the same time, we did not find a statistically significant result concerning a causal effect from political stability in energy consumption. Moreover, we checked the impact of the independent variables on the electricity consumption model, and finally, we investigated the causal relationships between these variables.

Apart from the introduction, the paper consists of four parts. Section 2 offers a theoretical background. Section 3 describes the econometric methodologies and the related data. Section 4 presents the empirical results and the related discussion, and Section 5 provides the concluding remarks and the policy implications.

2. Theoretical Background

2.1. The Impact of National Income and Economic Growth on Energy Consumption

The interaction of economic growth and energy demand and consumption affects the causal relationships between a country’s energy consumption level and the level of economic growth. Several studies have been conducted in this context, but not many of those concern a multivariate environment that results in consuming energy. One of the primary purposes of this research is to highlight the intensity of the interaction between the two quantities [38], and of course, the direction of causality. The dynamic econometric models used fully fulfil this purpose, taking into account past influences (FM-OLS, D-OLS) and extending them to the future (D-OLS). In addition, it is easy to include new control variables in the model to avoid the bias of the omitted variables, but caution is required to prevent the problem of multicolinearity [39].

The regression mentioned above has four possible directions [38,39,40]. The growth hypothesis refers to the unidirectional causality from energy consumption to economic growth, the reverse hypothesis of conservation, and the bidirectional causality that is the most common (feedback hypothesis) [41]. Finally, there is the hypothesis of neutrality, according to which economic growth and energy consumption do not causally interact. The present research fills a gap in the international literature through many different econometric processes. Results are extracted and confirmed with varying samples of country and various variables, and through causality tests according to Dumitrescu and Hurlin [26].

Ozturk and Acaravci [42] used average energy use, average electricity consumption, and GDP per capita to investigate the causality between energy and economic development in four southern European Countries throughout 1980–2006. The authors used the Engel–Granger method and found a two-way strong long-term association between energy utilization and GDP. However, a limitation of this study is that the author employed other methods such as autoregressive distributed lag (ARDL) model and found contradicting results showing no substantial relationship between the variables of interest. Furthermore, Mairet and Decellas [43] used the logarithmic mean division index I (LMDI I) decomposition method to assess energy demand in the service sector of France, using data from 1995–2006. The rise in energy utilization in the service sector was mainly due to structural economic growth.

The causes of energy efficacy and efficiency are measured by Sineviciene et al. [44] using the stochastic frontier analysis (SFA) to determine the causes of efficient energy consumption. The panel data were collected from 11 East European countries, post-communism from 1996 to 2003. The results indicate a positive relationship between increasing GDP per capita and energy efficiency. The study also estimated the positive correlation between industrial value-added and energy utilization. Conversely, higher technological export had a negative correlation with energy consumption. The same conclusions were reached by Sarkodie and Adom [45] for their research on businesses and households in Kenya.

Saldivia et al. [9], in their study, combine panel data techniques to investigate the relationship between energy consumption and GDP. The study takes into account heterogeneity and cross-sectional dependence. The analysis is applied in the 50 US states for the period from 1963 to 2017. It is then reproduced for a broad set of 25 subgroups of states, based on geography, income, energy intensity, energy price, and the economic sector. The results from the causality test show that in the short run, there are mixed data on the direction of causality between energy consumption and GDP. At the same time, there is a two-way causality for most of the subgroups in the medium and long term.

A specialized article by Rodríguez-Caballero [46] provides a process for evaluating large-panel data models with many data blocks subject to cross-sectional dependence. The model includes extensive dependency processes such as energy consumption and GDP, which means that the model can be applied to many other economic applications. The model is used to study the relationship between energy consumption and economic growth. The main findings suggest much lower elasticities between the two variables than those generally found in other empirical studies. Moreover, a key conclusion of the research is that causality moves mainly from energy consumption to GDP and less in the opposite direction. The proposed methodology proposes the review of causality studies between these two variables to clarify the impact of energy policies on economic development.

Finally, Lianos et al. [20] dealt with reverse causality. They studied the effects of energy consumption and carbon emissions on per capita economic growth with unbalanced panel data for 94 countries between 1971 and 2018. They assessed the effects using the potential dynamic outcomes under a framework of treatment-based causality framework. The results showed that the processing range that produces the best efficiency for both energy consumption and carbon emission policies is between −2% and 0.4%. In addition, in both cases, extreme policies such as drastically reducing or increasing energy consumption produce the worst results for economic growth.

From the above analysis, it appears that there is a gap in the literature regarding the study of multifactorial systems that estimate energy and electricity consumption. We aim to fill this gap with the present research.

2.2. The Impact of Financial Development on Energy Consumption

A healthy and well-developed financial system provides more financing and renewable energy sources at lower costs to the energy industry. The result is more investments that, in turn, boost energy demand again. Developed capital markets ensure that liquidity risk is reduced for energy production companies and help to raise the capital needed to develop energy-efficient technologies in the long run [18,47,48]. In addition, economic growth can facilitate the redistribution of capital from traditional energy with low production efficiency to renewable energy sources. This is likely to be the case for developed countries, as their inhabitants are more inclined to switch to low energy or fewer emissions but economically more expensive energy sources [49,50]. The impact of long-term financing on the consumption of renewable energy sources and the positive prospects for sustainable development justify the conduct of the present study on the relationship between economic growth and energy consumption, which is called the promoting effect [51,52,53].

However, there is also the opposite view. Some researchers claim there is a negative relationship between economic growth and energy consumption (inhibitory effect), and economic growth discourages energy consumption, which leads to lower pollution [54,55].

Sadorsky [13], in his study, evaluated the association between financial development and energy demand. The author used the mean group (MG) estimation in the model to specify the dependency of energy demand on income, price, and the measure of financial independence with panel data from 22 developing countries. The data ranged from 1990 to 2006. The study suggested stock market variables to be the most robust measure of financial growth that drives the energy demand and ultimately increases energy consumption. This paper has great contributions in expanding the field area and studying new relationships between the financial growth of a country and its energy demands.

Similarly, Shahbaz and Lean [56] found a positive correlation between energy utilization, GDP, financial development, industrialism, and urban expansion in Tunisia. The author collected data from 1971 to 2008 and employed the autoregressive distributed lag ARDL approach to identify correlation. The results indicated a direct association between financial growth and energy utilization in Tunisia, where a 10% increase in national lending would increase the demand for energy by 1.4%. Further on, Granger causality tests are used for the analysis.

Conversely, Topcu and Payne [57] evaluated the correspondence between energy consumption and financial growth in 32 countries throughout 1990–2014. The most significant aspect of this study is the indicators of financial development, which include variables from the bond market, stock market, and banking sector to imply their effects on energy consumption. The estimation results indicated an unsubstantial relationship between the financial development index and energy consumption in these countries. However, the variables from the stock market have a negatively substantial effect on energy utilization. The implications of this study suggested that energy-conserving policies would not affect the countries’ financial growth, and thus the energy consumption behavior will not change.

Another study assessed the link between financial development and energy use in Saudi Arabia. Mahalik et al. [58] used the data from 1971 to 2011 of economic development and urbanization as endogenous variables to determine their impact on energy demand. The results indicated a long-run relationship between financial development and energy demand, but economic growth had no significant effect on energy utilization. Lastly, a one-way relationship between financial growth and the country’s energy consumption was deduced. Finally, Mukhtarov et al. [59] estimated the relationship between financial development, economic growth, and energy expenditure. The author used cointegration techniques on data ranging from 1992 to 2015. The findings indicated a substantial correlation between financial growth, economic development, and energy consumption in the long run.

The study by Chiu and Lee [60] explored the impacts of countries’ risks on the relationship between energy consumption and financial development for 79 countries from 1984 to 2015. They showed that banking sector development has more significant impacts on energy consumption than stock market development. The total sample results showed that financial development could help to reduce energy consumption. Lastly, the results showed that different types of financial development and country risk environments have varying impacts on energy consumption in OECD and non-OECD countries.

From the above literature, it seems that no study examines the recent data to show how financial growth affects energy consumption until 2018. This gap is trying to fill the current research to conclude whether the financial factor affects today’s energy price crisis.

2.3. Direct and Indirect Effect of Corruption on Energy Consumption

According to the literature, corruption can affect energy consumption in various ways. First, greater corruption reduces energy policy rigor [30], increasing energy consumption. Second, the growing cost of coordinating bribery leads to a stricter energy policy [61]. Third, growing corruption is causing a reduction in per capita output while reducing energy consumption. Various studies consider corruption a central factor affecting energy consumption and environmental quality, where corruption and environmental problems have been significant challenges to sustainable economic development. The present study tries to shed light on this interaction and determine whether the final effect of corruption is positive or negative and whether this relationship is causal.

The study by Fredriksson et al. [30] theoretically indicated that corruption in government institutions and lobby groups in energy-intensive industries tend to reduce the effectiveness of energy policies. The panel data in this paper were collected from 11 sectors in 12 Organisation for Economic Co-operation and Development (OECD) countries and ranged from 1982 to 1996. The model estimated the causal relationship and evaluated the impact of rent-seeking behavior of union groups in the industries and government institutions where these variables render the energy policy ineffective. One limitation of this paper is the lack of evidence to negate the idea that corruption might be affected by energy policy, which can lead to endogeneity in the model. However, this paper offers profound policy implications in OECD countries to reduce the corruption index.

Sekrafi and Sghaier [31] indicated a significantly negative relationship between corruption control and energy consumption in the Middle East and North Africa (MENA) region. The paper studied this relationship through economic growth as a proxy of energy consumption. The data are collected from WDI ranging from 1984 to 2012. Energy consumption is positively correlated with economic growth, where a one-unit increase in GDP results in 1.6 more units of energy demanded.

Arminen and Menegaki [5] estimated the interconnection between economic growth, emissions, and energy consumption from a group of developed and developing countries using the simultaneous equation modeling through the lens of institutional quality and corruption. The results indicated a direct relationship between average energy consumed and average income, which also justified the feedback hypothesis of the study. More importantly, corruption did not appear to be a strong determinant of energy consumption and CO₂ emissions.

In their theoretical study, Boamah et al. [10] found evidence that Africa’s energy sector relies on corrupt practices to tackle entrenched energy injustices. In Kenya, unequal spatial distribution of electricity networks, bureaucracy, “unfair” electricity billing systems, and lower energy production of decentralized solar photovoltaic systems have forced energy users to resort to corruption to increase the speed of access to the grid. Similar processes have been observed in Ghana. Their article presents research data from Ghana and Kenya that show that energy injustice and corruption are inherent in both countries’ “power regimes” and that corruption in these cases can also lead to increased electricity consumption.

There seems to be conflicting conclusions about the impact of corruption on energy consumption. In the present study, we use corruption in all models and sub-models (developing and developed countries, electricity, and energy consumption) to clarify its correlation with energy consumption.

2.4. The Effect of Investment, Innovation, Environmental, and Energy Distribution Factors on Energy Consumption

In addition to the abovementioned factors that theoretically affect energy consumption, several others need to be considered in a study that seeks to create a holistic and multifactorial model. The main variables are investments, innovation aimed mainly at renewable energy sources, the impact of the climate, and losses in the energy production and distribution system.

Hao et al. [62] estimated energy utilization, investment, and economic development in rural areas using a panel dataset from 1995 to 2010 for 28 provinces of China. The results indicated a positive and bidirectional interconnection between rural GDP and rural energy use in the short run. In contrast, a bidirectional positive relationship exists between rural investment and rural economic growth in the long run. Moreover, it is observed that rural energy consumption and rural investment positively impacted rural GDP in the long term.

In addition, Amin et al. [11] assessed the effects of government and private funding on the level of energy utilization in South Asian countries. The study is significant as it fills a gap in this field area. The authors used a panel framework throughout 1980–2016. The cointegration between variables is identified through the Durbin–Hausman tests. The findings suggested no correlation between government financing and energy utilization; however, increasing the private investment by 1% increases energy consumption by 0.35% in the long term.

Similarly, Ugursal [63] presented increasing energy consumption due to economic and human development indicators and innovation. The author analyzed the research question qualitatively and found that energy consumption can only be reduced in developed countries without reducing the Human Development Index (HDI). In contrast, the developing country can only benefit from a minimum increase in energy consumption. Lastly, the author evaluated that developed countries must meet a minimum level of energy consumption for improved socio-economic conditions of a country. Therefore, it is concluded that the growing innovation in technology will escalate the consumption of energy and ultimately affect the socio-economic conditions of the population.

Latief et al. [37], in their study, analyzed the relationship between economic growth, energy consumption, and sustainable development. They used a sample of 14 developed and developing Mediterranean countries, using the generalized method of moments. Moreover, they used causality analysis to examine the long-term and short-term causal relationship between their variables. The results confirmed the short-term dynamic correlation from sustainable growth to energy consumption and economic growth to sustainable development.

Alfredsson [64] evaluated the impact of renewable energy on energy consumption levels and patterns along with greenhouse gas emissions. The author used data from 1140 households’ consumption levels that were categorized into 300 subjects, and utilized a microsimulation model. The results indicated that a complete shift to renewable energy sources would only change the energy consumption levels to insignificantly minimum levels. In addition, the study by Liu et al. [65] addressed the influence of energy efficiency policies and regulations on patterns of energy consumption. The authors used data from 1997 to 2015 from China and found that the overall effect of environmental regulations in energy policies related to energy consumption is harmful.

Likewise, Gielen et al. [66] evaluated the contribution of green energy sources in total energy utilization by 2050. The study indicated certain essential factors in considering the transition towards renewable energy. The technology of renewable energy sources, energy efficiency, and contribution towards total energy consumption are crucial variables to consider in shifting from non-renewable energy consumption to clean and green energy sources. Renewable energy sources can only contribute to 1/3rd of the total energy demand. Therefore, a comprehensive policy for clean energy must be devised to contribute to total energy consumption.

The existing literature on the energy–environment–economic growth relationship has not considered the impact of climate on energy consumption. However, weather fluctuations can significantly affect energy demand. An advantage of using geographic factors as explanatory variables is that they are exogenous. In a recent study, Stern et al. [67] assessed the factors driving the long-term growth rates of per capita emissions of air pollutants. They included the average temperatures for three summer and winter months in their model and found that temperature changes were statistically significant emissions determinants in most models. In addition, Stern and van Dijk [34] examined the effect of climate by including average temperatures during the summer and winter months when analyzing how economic growth affects air pollution. Their results show that air pollution is higher in countries with lower average winter temperatures and higher summer temperatures. Arminen and Menegaki [5] showed that the average temperature for three winter months strongly affects energy consumption. It seems that increased heating requirements are strongly associated with higher energy demand in high- and upper-middle-income countries.

The studies mentioned above indicate that weather variations play a significant role in energy consumption. We, therefore, consider it necessary to incorporate these variables into our overall model.

On the impact of consumption of electricity utilization and transmission losses on the economic growth, Oseni [33] found that the provision of modern energy sources increases the economic development in the country and brings sustainability. The author presented the case study of Nigeria, where 40% of the population does not have electricity access and relies on traditional methods. The author found significant energy distribution losses due to poor equipment as the data showed 4.49% lower electricity distribution in 2008 compared to 2007. Additionally, Tan et al. [68] explained that the increasing demand for energy consumption was due to residents’ higher levels and earnings. Moreover, it was presented that the ratio of electricity consumption to the total energy mixes consumption increased from 17.4% to 21.7% in just five years. Therefore, it is argued that the level of electricity consumption is rising along with the rise in total energy consumption level. Finally, in their theoretical study, Pavlicko et al. [69] proposed forecasting models to predict the maximum hourly electricity consumption per day of the Slovak Distribution Company. Many different models have been proposed: gray and nonlinear gray Bernoulli models, models based on a multi-level front-to-rear feed network, and a new hybrid model that combines these different approaches. All the proposed models achieved more accurate predictions than the official load forecast. In contrast, the hybrid model offered the best results, essentially offering a standard shape through which it is possible to predict electricity consumption.

From the research presented, it seems that the treatment of energy consumption in terms of its causes was fragmentary and data were from small samples of countries, in contrast to that of carbon dioxide emissions. The present study attempts to supplement the existing literature using a large number of determinants (11), a large sample of countries (109), and many reliable econometric methodologies (4).

Table 1. Selected empirical literature on the relationship between energy and electricity consumption and their determinants.

Author	Countries or Territories	Period	Main Objective	Method	Major Findings
Sadorsky [13]	22 emerging economies	1990–2006	The effect of financial development on energy consumption	GMM	Financial development has a positive and significant impact on energy consumption.
Arminen and Menegaki [5]	67 Countries	1985–2011	The effect of corruption and climate on the energy–environment–growth nexus	Simultaneous equations model	Climate plays a more significant role in energy consumption and emissions than corruption.
Hajko et al. [6]			The energy–growth nexus: history, development and new challenges	Theoretical research	Review of the energy-growth literature.
Mudakkar et al. [7]	5 SAARC countries	1975–2011	The investigation of the multivariate energy consumption function	Granger causality tests	Causalities in the five countries between FDI, GDP, energy consumption, financial dev. and energy prices.
Flores-Chamba et al. [12]	34 European countries	2000–2016	The economic and spatial determinants of energy consumption in the European Union	Panel data, spatial Durbin model (SDM)	The price of oil and human capital have a negative effect on energy consumption.
Khanna and Rao [14]	Developing countries		Supply and demand of electricity factors in the developing world	Theoretical	The electricity demand is driven by GDP, prices, income, level and characteristics of economic activity and urbanization, and seasonal factors.
Lee and Chang [15]	16 Asian countries	1971–2002	The co-movement and the causal relationship between energy consumption and real GDP	Cointegration and panel error correction models	There is a long-run unidirectional causality running from energy consumption to economic growth.
Khan et al. [18]	Pakistan	1971–2016	The impact of globalization, economic factors, and energy consumption on CO2 emissions	The dynamic ARDL simulations model	Economic, social, and political globalization have positive effects, and urbanization, economic growth, and innovation harm CO2 emissions.
Anton and Nucu [22]	28 European Union countries	1990–2015	The effect of financial development on renewable energy consumption	Panel fixed effects model	Financial development has a positive effect on the share of renewable energy consumption.
Khan et al. [25]	184 countries	1990–2017	Impact of financial development and energy consumption on environmental degradation	GMM, seemingly unrelated regression-SUR	Economic growth has a positive effect on CO2 emissions.
Sekrafi and Sghaier [31]	13 MENA countries	1984–2012	The effect of corruption, economic growth, and environmental degradation on energy consumption	Fixed and random effects, GMM	Corruption and economic growth positively affect CO2 emissions and energy consumption.
Vasylieva et al. [32]	28 European countries and Ukraine	1996–2007	The dynamic impact of renewable energy consumption, GDP, and corruption on gas emission	FMOLS, DOLS	Renewable energy consumption and corruption negatively affect GHG emissions.
Oseni [33]	Nigeria	1990–2008	Improving households’ access to electricity and energy consumption pattern in Nigeria	Comparative research	The access to the modern form of energy in the country is deficient despite the country’s abundant energy endowment.
Payne [38]	56 countries		Survey of the causal relationship between energy consumption and growth	Survey of the literature	29.2 percent of the results supported the neutrality hypothesis; 28.2 percent the feedback hypothesis; 23.1 percent the growth hypothesis; and 19.5 percent the conservation hypothesis.
Ozturk and Acaravci [42]	4 countries	1980–2006	The causal relationship between energy consumption and GDP in Albania, Bulgaria, Hungary, and Romania	ARDL, vector error correction (VEC)	There is a long-term relationship between energy consumption variables and real GDP per capita only in Hungary.
Sineviciene et al. [44]	11 East Europ. countries	1996–2003	Determinants of energy efficiency and energy consumption of Eastern Europe economies	Stochastic frontier analysis (SFA)	The results indicate a positive relationship between increasing GDP per capita and energy efficiency.
Sarkodie and Adom [45]	Kenya	1971–2014	Determinants of energy consumption in Kenya	Nonlinear iterative partial least squares method	Renewable energy reduces total energy consumption. Other renewable energies increase total energy and electricity consumption.
Saldivia et al. [9]	50 US states	1963–2017	The relationship between energy consumption and GDP	Panel data and wavelet spectral analysis	There are mixed data on the direction of causality between energy consumption and GDP in the short run, while there is a two-way causality in the medium and long term.
Rodríguez-Caballero [46]			A theoretical study of the nexus between energy consumption and economic growth	Block-by-block cross-sectional averages	Causality moves mainly from energy consumption to GDP and less in the opposite direction.
Lianos et al. [20]	94 countries	1971–2018	The effects of energy consumption and carbon emissions on per capita economic growth	Causality framework	Extreme policies such as drastically reducing or increasing energy consumption produce the worst results for economic growth.
Wang et al. [48]	11 countries (emerging markets)	1990–2017	The determinants of CO2 emissions in N-11 countries	Pesaran’s unit root test	Renewable energy consumption, innovation, and human capital development reduce CO2 emissions.
Topcu and Payne [57]	32 countries	1990–2014	The correlation between energy consumption and financial growth	Principal component analysis	Energy-conserving policies would not affect the countries’ financial growth.
Mahalik et al. [58]	Saudi Arabia	1970–2011	The relationship between financial development and energy use	ARDL and combined cointegration tests	A long-run relationship between financial development and energy demand is found, but economic growth has no significant effect on energy utilization.
Chiu and Lee [60]	79 countries	1984–2015	The impacts of country risks on the relationship between energy consumption and financial development	Panel smooth transition regression model	The banking sector development has more significant impacts on energy consumption than stock market development.
Boamah et al. [10]	Kenya and Ghana	2017–2019	The relationship between corruption and the electricity sector	Comparative research	Energy injustice and corruption are inherent in both countries’ “power regimes,” corruption can also lead to increased electricity consumption.
Hao et al. [62]	28 provinces of China	1995–2010	The relationship between energy consumption, investment, and economic growth	FMOLS, Vector error correction model	There is a positive and bidirectional relationship between rural GDP and rural energy use in the short run, whereas a bidirectional positive relationship is shown in the long run.
Amin et al. [11]	South Asia countries	1980–2016	The government versus private investment on energy consumption in South Asia	Cointegration techniques	The findings suggested that increasing the private investment by 1% increases energy consumption by 0.35% in the long term.
Latief et al. [37]	14 Mediterranean countries	1995–2014	The relationship between economic growth, energy consumption, and sustainable development	GMM and causality analysis	The results confirmed the short-term dynamic correlation from sustainable growth to energy consumption and economic growth to sustainable development.
Gielen et al. [66]			The contribution of green energy sources in total energy utilization	Theoretical research	Renewable energy sources only contribute to 1/3rd of the total energy demand.
Pavlicko et al. [69]	Slovakia	2000–2020	Forecasting models to predict the maximum hourly electricity consumption per day of the Slovak Distribution Company	Theoretical study	They considered a model which offered a standard shape through which it is possible to predict electricity consumption.

presents the summary of the significant and most recent previous studies.

3. Research Methodology and Data

Our goal is to improve the current literature on the factors that affect energy consumption. For this reason, we include all the key factors that affect it in an equation to arrive at a complete model that describes the phenomenon. The equation of the base model was transformed and derived indirectly from the determinants as mentioned above. Following Arminen and Menegaki [5], our model was specified as follows:

$${\mathrm{ec}}_{\mathrm{it}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1 {\mathrm{cor}}_{\mathrm{it}}+{{\displaystyle \sum }}_{j=2}^m{\mathsf{\beta }}_j{x}_{\mathrm{it}}^j+{\mathsf{\epsilon }}_{\mathrm{it}},$$

(1)

where ec_it is energy or electricity consumption, cor_it is the corruption indicator, $x_{\mathrm{it}}^j$ are m − 1 explanatory variables that can affect energy consumption without being particularly associated with the corruption, namely real GDP per capita, the average temperatures over three winter months, energy losses, innovation, trade openness, total investment, population growth, political stability, the financial development, and the renewable energy production. ε_it is the error term. Following the most relevant previously mentioned studies, energy consumption is given a function of many causes as follows:

LENC_it = β₀ + β₁ LCPI_it + β₂ LGDP_it + β₃ ENRLOS_it + β₄ RENW_it + β₅ PS_it + β₆ INVT_it + β₇ POP_it + β₈ TEMPWIN_it + β₉ TROP_it + β₁₀ FINDEV_it + β₁₁ RDEXP_it + ε_it,

(2)

LELC_it = β₀ + β₁ LCPI_it + β₂ LGDP_it + β₃ ENRLOS_it + β₄ RENW_it + β₅ PS_it + β₆ INVT_it + β₇ POP_it + β₈ TEMPWIN_it + β₉ TROP_it + β₁₀ FINDEV_it + β₁₁ RDEXP_it + ε_it,

(3)

3.1. Research Framework

The primary purpose of the research is to assess the factors that affect energy consumption and whether their impact is different in different groups of countries.

After the factors were identified in the introduction and the literature review sections, the equations were created through which the model will be empirically investigated. The expected outcomes (signs) of the independent variables on energy consumption based on the literature are depicted in

Table 2. Direction of the regression with the dependent variable (energy consumption).

Variable	Sign	Variable	Sign	Variable	Sign
Corruption	+ or neutral	LGDP	+ or −	Energy losses	+
Renewables	−	Political stab.	− or neutral	Investment	+
Pop. growth	+	Winter temp.	−	Trade openness	+
Financial development	+ or neutral

.

As several macro variables are used, and there is a possibility of cross-sectional dependence (CSD) between them, the corresponding tests must be performed.

Appropriate econometric techniques must be used to address the problems of endogeneity, heteroskedasticity, and, especially, the CSD. These methodologies are the error correction (PCSE, PCSE, and AR1) [27], the dynamic (FM-OLS and D-OLS) [28,29], and the GMM [28]. Because the application of GMM creates problems related to interpreting the lag variables, which are analyzed below, we use the error correction and the dynamic methodologies. The steps to apply these techniques are as follows:

(a)

The unit-roots are checked;

(b)

The CSD tests are performed;

(c)

The panel cointegration test is performed.

Next, we estimate Equation (2) coefficients with the techniques mentioned above.

We check our conclusions by comparing them with those in Table 2.

Our primary independent variable is energy consumption, but electricity consumption is also used, which is of a different form to confirm our basic conclusions (Equation (3)).

At the same time, for more information and a more significant robustness check, we divide the sample into developing and developed countries and rerun the basic regressions.

Finally, we apply causality tests to check for causal relationships between independent and dependent variables.

3.2. Econometric Methodologies

The probability of omitted variables, heterogeneity, and multicollinearity issues led us to perform various tests: the t-test, Breusch–Pagan/Cook–Weisberg test, Ramsey test, and F-test. The Breusch–Pagan/Cook–Weisberg test showed a problem with heterogeneity in the country data. By applying the Wooldridge test for autocorrelation, the modified Wald method, and the Breusch Pagan test for heteroscedasticity and identifying the issues of autocorrelation and heteroscedasticity, we introduced error correction models with panel corrected standard errors (PCSE and PCSE (AR1)) [70].

Endogeneity is particularly problematic in empirical research related to corruption and energy consumption. To address endogeneity, we applied FM-OLS and D-OLS methodologies. According to Pedroni [71], FM-OLS addresses the problem of both endogenous and omitted variables effectively. It is challenging to identify an auxiliary variable associated with the abovementioned variables rather than the error term. Due to this phenomenon, the two-stage least squares (2SLS) methodology has weak auxiliary variables. On the other hand, FM-OLS uses semi-parametric corrections in the estimators of the OLS methodology to rule out second-order problems due to the endogenous nature of the independent variables.

These procedures are increasingly used in the literature, as the corresponding GMMs have, in addition, many disadvantages in our holistic model. First, the large number of choices from instruments and variables leads to an overestimation of the coefficients of endogenous variables [72]. Second, it is assumed that the coefficients and their variations are homogeneous within the panels, which leads to inconsistencies in the estimates of the dynamic samples [73]. Finally, the difficulty of locating the appropriate tools shows the tendency in which the time lag of the dependent variable occupies a considerable part of the interpretation of the equation [74]. The result is that the other variables have to interpret a tiny amount of the model under consideration, which has a cost in terms of pluralism and substance of the research.

We first determined whether the variables included in the model had stationarity. Because the variables under investigation were cointegrated, we estimated their relationship in the long-run equilibrium state. When applied to cointegrated panels, the least-squares method (OLS) yielded discriminatory estimators, especially when the sample size relative to periods (t) was small. Therefore, to calculate valid t-statistics, we estimated the long-run equilibrium function with the FM-OLS procedure proposed by Pedroni [71] and the D-OLS proposed by Stock and Watson [75]. These approaches provide more reliable results when heterogeneity in integrated variables is diagnosed.

3.3. Data and Descriptive Statistics

We used a panel dataset from 109 countries over 2010–2018. Specifically, the countries were used in maximum numbers regarding reliable data for some variables, such as energy losses, patents, and energy consumption. Finally, the primary purpose was to represent all regions internationally so that our conclusions apply generally. The first region comprises 40 Europe and Central Asia (ECA) countries. The second consists of 17 Latin American and Caribbean (LAC) countries, the third of 14 East Asia and Pacific (EAP) countries, the fourth of 9 Middle East and North Africa (MENA) countries, and the fifth of 24 Sub Saharian and African (SSA) countries. There are also 2 North American (NA) and 3 South Asian (SA) countries. The separation has been made according to the World Bank. The time was chosen based on the post-financial crisis period to determine the correlation of the variables and their agreement with the literature thus far and to draw helpful research and policy conclusions.

3.4. Criteria Selection

A multivariate model was used to determine energy and electricity consumption in the present study. Because of the context of using specific variables, the criteria by which they have been selected needs to be explained. There are many determinants for energy consumption. To have uncorrelated variables, we chose the basic ones and one of each category of the causes.

Economic growth has resulted in increased energy consumption. The critical variable of economic growth is considered to be GDP per capita so that it is comparable from country to country [18,19]. The second primary reason for the increase in energy consumption is financial growth. Increasing the circulation of money and all financial products leads to growth leading to energy consumption. The variable used here is financial development and has been previously used extensively in the literature [13,57]. Investments in the whole economy would be required to upgrade the infrastructure of production, transport, and energy usage. At the same time, investment boosts economic activity and increases energy demand. The variable used here is cross capital formation (% GDP) and is a crucial measure of investment by the World Bank [11,62]. Innovation is also a key determinant of energy consumption. The innovative practices in the entire range of production, distribution, and usage are significant causes of energy consumption while they also replace the outdated systems in an innovative way; thus, increasing the efficiency of the overall process [24,25]. Three different indicators have been used for robustness. The GII was calculated through surveys and the number of patents per million inhabitants and the R&D expenses are highly realistic innovation measurements. These three different measurements give an overall picture of innovation.

Another factor that contributes to the energy crisis is the global population and its demands for fuel and products. All types of production of food and other products and resources use energy during their production and transportation. As countries are affected when it comes to consumption and labor force, it is required to use these variables in the model. The basic notion is that the variables work differently in developing than developed countries. The proxy of this variable is population growth [18,23]. At the same time, the overuse of natural resources deepens the problem. As mentioned earlier, renewable energy sources are still unused in most countries, and their use can reduce dependence on fossil fuels and help reduce greenhouse gas emissions [22]. The variable used is renewable energy consumption (% of final energy consumption) [65]. In addition, an essential factor is the ageing of power generation equipment infrastructure. The more the facilities and transmission lines experience wear and tear, the more energy will be required. The variable used is electric power transmission and distribution losses (% of output) so that it is comparable from country to country [76].

The extreme temperature climatic factor was also used to consider the model in terms of energy consumption [5,34]. We chose winter temperature (average temperatures over three winter months: December, January, and February) to avoid the correlations of the independent variables. Political instability is also a critical factor in energy demand. Internal and external imbalances can cause changes in the energy requirements from one country to the other. The variable used for this is political stability [77]. In many countries, there is a significant delay in launching new power plants that can bridge the gap between energy supply and demand, and a delay in issuing licenses for renewable energy sources. A key reason behind this is the bureaucracy and corruption inherent to such structures, which will be the variable used in the present study [31,32]. We used many corruption variables (CPI, CCI and ICRG) to strictly control the results as the data in the existing literature are ambiguous, as mentioned above.

Finally, for energy consumption, the variable energy consumption per capita was used to measure per capita consumption as it is comparable from country to country. In our model, we try to reach conclusions about the factors that change energy consumption as well as compare them from country to country. Electricity consumption per capita has been used for a robustness check.

Other factors that affect energy consumption were omitted from this study as they either were of little importance (such as urbanity or over-taxation) or because others were used that shared a tight relationship with them (such as economic complexity, GDP, and trade openness). Some were also omitted due to insufficient data (economic policy uncertainty). Specifically, the Economic Complexity Index ranks countries based on the diversity and complexity of their export basket. High complexity countries are home to a range of sophisticated and specialized capabilities and can produce a highly diversified set of complex products. Moreover, economic complexity is related to a country’s level of prosperity, and there is a tight relationship between economic complexity and GDP per capita [78]. The Economic Policy Uncertainty Index by Baker et al. [79] takes into account only a limited number of countries.

3.5. Energy and Electricity Consumption, Energy Losses, and Renewable Energy Consumption

Energy use refers to the use of primary energy before transformation to other end-use fuels, equal to indigenous production plus imports and stock changes, minus exports and fuels supplied to ships and aircrafts engaged in international transport. The data were obtained from the World Bank database [80]. The energy consumption was calculated as a kg of oil equivalent per capita. The electricity consumption was calculated as kWh per capita. Electric power consumption measures the production of power plants and combined heat and power plants less transmission, distribution, and transformation losses, and own use by heat and power plants [80]. The data were obtained from the World Bank database.

The energy losses were calculated as electric power transmission and distribution losses (% of output). These include losses in transmission between sources of supply and points of distribution and in the distribution to consumers, including pilferage [80]. The renewable energy consumption was calculated as a % of total final energy consumption. All the data were obtained from the World Bank database.

3.6. Average Temperatures over the Three Winter Months

The average temperatures over the three winter months capture the effect of climate and weather variations [5,34,67]. The lower the average winter temperature, the higher the heading requirement, and the higher the resulting energy consumption is expected to be. We could report the same for the summer months, but we used only one variable to avoid multilinearity in the model. We used the Climate Change Knowledge Portal (World Bank) data, as this portal is committed to transparency and data availability. Next, we used monthly data for each year and each country. Finally, we calculated the average quarterly temperature for the winter months: December, January, and February. There are data from 1901 to 2020 (WB Group, Climate Change Knowledge Portal).

Table 3. Variables, Symbols, Literature, Data descriptions, and Sources.

Symbols	Variables	Time	Data Sources	Literature
LENC	Natural logarithm of energy consumption (kg of oil equivalent per capita)	2010–2018	WB, 2021	Anton and Nucu [22]; Wang et al. [48]
LELC	Natural logarithm of electricity consumption (kWh per capita)	2010–2018	WB, 2021	Oseni [33]; Tan et al. [68]
RDEXP	Research and development expenditure (% of GDP)	2010–2018	WB, 2021	Balsalobre-Lorente et al. [81]
ENRGLOS	Electric power transmission and distribution losses (% of output)	2010–2018	WB, 2021	Shahbaz et al. [76]; Oseni [33]
LGDP	GDP per capita at fixed USD prices (2011) (PPP)	2010–2018	WB 2021	Sineviciene et al. [44] Ozturk [39]
FINDEV	Financial development: domestic credit to the private sector (% of GDP)	2010–2018	WB, 2021	Topcu and Payne [57]; Sadorsky [13]
TROP	Total exports and imports as a percentage of a country’s GDP, at constant prices	2010–2018	Global Economy 2020	Khan et al. [25]; Nasreen and Anwar [82]
TEMPWIN	Average temperatures over three winter months (Dec., Jan., Feb.)	2010–2018	WB Group, CCKP, 2022	Arminen and Menegaki [5]; Stern and van Dijk [34]
RENW	Renewable energy consumption (% of final energy consumption)	2010–2018	WB, 2022	Liu et al. [65]; Alfredsson [64]
LCPI	Natural logarithm of Corruption Perception Index, 0–100 (0 low)	2010–2018	Transparency International	Fredriksson et al. [30]
LCCI	Control of Corruption Index (−2.5–2.5, transformed in 0–100, rev)	2010–2018	WB (WDI), 2021	Sekrafi and Sghaier [31]
ICRG	International Country Risk Guide (Corruption Index) (0–6)	2010–2018	Transparency International	Arminen and Menegaki [5]
INVT	Total investments (% GDP)	2010–2018	World Bank, Global Econ.	Amin et al. [11]; Hao et al. [62]
POP	Population growth (annual %)	2010–2018	WB, 2021	Khan et al. [18]; Zaman et al. [23]
LGII	Natural logarithm of Global Innovation Index (0–100)	2010–2018	INSEAD and World Busin.	Ellis et al. [83]
PAT	Patent applications, all (per capita, per million)	2010–2018	WB, 2021	Ellis et al. [83]; Dincer [74]
PS	Political stability (−2.5–2.5)	2010–2018	WB (WDI), 2021	Cieślik and Goczek [77]

shows all the variables used in the paper, with the symbols, the sources, and the corresponding bibliography.

4. Empirical Results and Discussion

In Appendix A, we present the statistical characteristics of the variables in a table with the total sample.

4.1. Preliminary Tests of Econometric Analysis (FMOLS Methodology)

4.1.1. Unit Root Tests

We used unit root stationary tests introduced by (a) Levin, Lin, and Chu, (b) Breitung, (c) Maddala and Wu, and Choi (augmented Dickey–Fuller test (ADF) and Fischer).

Table A2. LLC, BREITUNG, and ADF–FISHER unit roots test at levels and first differences.

Variables	LLC		BREITUNG		ADF–FISHER
	Levels	1st Difference	Levels	1st Difference	Levels	1st Difference
LENC	−10.178 *** (0.000)	−36.634 *** (0.000)	4.841 (1.000)	−8.986 *** (0.000)	−1.151 (0.125)	−19.037 *** (0.000)
LELC	−17.419 *** (0.000)	−22.966 *** (0.000)	6.151 (1.000)	−9.989 *** (0.000)	−1.100 (0.136)	−21.926 *** (0.000)
RDEXP	−3358 *** (0.000)	−26.839 *** (0.000)	4.518 (1.000)	−9.450 *** (0.000)	3.582 (0.999)	−17.635 *** (0.000)
ENRGLOS	0.315 (0.624)	−1.4 × 107 (1.000)	1.242 (0.893)	−7.688 *** (0.000)	−6.994 *** (0.000)	−33.385 *** (0.000)
LGDP	−6.805 *** (0.0031)	−23.614 *** (0.000)	14.911 (1.000)	−3.060 *** (0.001)	2.983 (0.999)	−17.019 *** (0.000)
FINDEV	−10.271 *** (0.000)	−20.438 *** (0.000)	8.177 (1.000)	−5.776 *** (0.000)	0.609 (0.729)	−10.775 *** (0.000)
TROP	−10.895 *** (0.000)	−14.157*** (0.000)	1.993 (0.977)	−6.844 *** (0.000)	−1.736 ** (0.042)	−20.652 *** (0.000)
TEMPWIN	−10.242 *** (0.002)	−24.845 *** (0.000)	−6.888 *** (0.000)	−8.933 *** (0.000)	−23.345 *** (0.000)	−38.454 *** (0.000)
RENW	−24.729 *** (0.000)	−31.659 *** (0.000)	5.306 (1.000)	−9.53*** (0.000)	2.828 (0.998)	−16.647 *** (0.000)
LCPI	−40.26 *** (0.000)	−25.190 *** (0.000)	−0.238 (0.406)	−12.417 *** (0.000)	−3.446 *** (0.000)	−27.282 *** (0.000)
LCCI	−16.053 *** (0.000)	−22.126 *** (0.000)	0.014 (0.506)	−10.791 *** (0.000)	0.264 (0.604)	−15.476 *** (0.000)
ICRG	−100.01 *** (0.000)	−35.391 *** (0.000)	1.431 (0.924)	−8.840 *** (0.000)	−4.638 *** (0.000)	−15.307 *** (0.000)
INVT	−18.171 *** (0.000)	−27.222 *** (0.000)	0.204 (0.581)	−7.841 *** (0.000)	−5.539 *** (0.000)	−18.768 *** (0.000)
POP	−15.516 *** (0.000)	−42.679 *** (0.000)	6.627 (1.000)	2.468 (0.993)	−10.382 *** (0.000)	−13.787 *** (0.000)
LGII	−26.237 *** (0.000)	−30.451 *** (0.000)	−1.292 * (0.098)	−14.832 *** (0.000)	−4.529 *** (0.000)	−21.361 *** (0.000)
TRMRK	−12.102 *** (0.000)	−7.720 *** (0.000)	5.910 (1.000)	−7.574 *** (0.000)	−9.870 *** (0.000)	−33.055 *** (0.000)
PAT	−6.243 *** (0.000)	−22.989 *** (0.000)	1.887 (0.970)	−9.463 *** (0.000)	−4.758 *** (0.000)	−26.733 *** (0.000)
PS	−10.667 *** (0.000)	−19.951 *** (0.000)	−1.106 (0.134)	−9.391 *** (0.000)	−3.356 *** (0.000)	−24.942 *** (0.000)

The table presents the panel and 1st difference unit root results for 109 countries. In all tests, the null hypothesis is that there is a non-stationarity. The numbers in parentheses indicate the p-values. *, **, and *** symbolize the rejection of the null hypothesis at a significance level of 10%, 5%, and 1%, respectively. Al abbreviations and sources are the same as in Table 3.

in the Appendix B shows the results, including values at the levels and the first differences. All variables were stationary in the first differences. The null hypothesis of non-stationarity was rejected at the 99% significance level for all tests.

4.1.2. Cross-Sectional Dependence (CSD) and Panel Cointegration Results

Next, we ran the CSD tests proposed by Pesaran and Frees, with models containing two, five, and seven variables. Cross-sectional dependence was confirmed for all models with 4, 8, and 12 variables, meaning that one country’s change is likely to affect another’s. To apply FMOLS econometrics, we adopted panel cointegration tests, proposed by Kao, Pedroni, and Westerlund, to identify cointegrated vectors in our key variables. The results showed cointegration between the model’s main variables, LENC as a dependent variable, and ENRGLOS, LGDP, RENW, PS, FINDEV, INV, and LCPI as regressors, forming a long-term equilibrium relationship.

4.2. Impact of the Determinants on Energy Consumption

The proposed analysis included initial correlations, calculated by Hausman, Woolridge, and modified Wald tests, which showed that the model had heteroscedasticity, autocorrelation, and cross-sectoral dependence. The PSCE and PSCE (AR1) models were used for correction, as shown in

Table 4. Main results in the full model with error correction and FMOLS estimation method.

Dependent Variable	Energy Consumption (LENC) 981 Observations, 109 Countries (Total)
Model	PCSE	PCSE (ar1)	FMOLS (4)	FMOLS (8)	FMOLS (12)	DOLS
ENRGLOS	−0.004940 *** (0.000)	−0.002713 (0.105)	−0.014522 *** (0.003)	−0.006728 *** (0.003)	−0.002462 (0.228)	−0.00426 (0.347)
LGDP	0.647754 *** (0.000)	0.610415 *** (0.000)	0.725809 *** (0.003)	0.602078 *** (0.000)	0.61438*** (0.000)	0.65234 *** (0.000)
RENW	0.004682 *** (0.000)	0.002529 *** (0.004)	0.005422 *** (0.002)	0.003199 *** (0.000)	0.003487 *** (0.000)	0.00569 *** (0.002)
PS	0.022503 (0.124)	0.003681 (0.799)		0.019460 (0.412)	0.092109 *** (0.000)	0.02778 (0.616)
FINDEV	0.001053 *** (0.000)	0.000939 *** (0.007)		0.004738 *** (0.000)	0.003620 *** (0.000)	0.00114 (0.287)
INVT	−0.007409 (0.103)	−0.001893 * (0.097)		−0.012577 *** (0.000)	−0.021518 *** (0.000)	−0.00767 (0.138)
POP	0.101719 *** (0.000)	0.034431 * (0.068)		0.034596 ** (0.030)	0.223319 *** (0.000)	0.09567 ** (0.010)
LCPI	0.164759 *** (0.000)	0.010045 (0.747)			0.432699*** (0.000)	0.21773* (0.067)
TROP	0.000827 *** (0.000)	0.000346 (0.110)			0.000556 ** (0.037)	0.00114 ** (0.048)
TEMPWIN	−0.022503 *** (0.000)	−0.015637 *** (0.000)			−0.021967 *** (0.000)	−0.02277 *** (0.000)
RDEXP	0.104405 *** (0.000)	0.076072 *** (0.000)			0.198643 *** (0.000)	0.11945 ** (0.018)
$R^2$	0.8563	0.9899	0.2381	0.2321	0.2941	0.8656
Hausman Test	Chi2(11) = −294.69
Wooldridge Test for auto/relation	F(1108) = 130.567 ***
Modified Wald test for gr/pwise heteroscedasticity	Chi2(109) = 12801.05 ***

*, **, and *** mean the significance level of 10%, 5%, and 1%, respectively, the numbers in parentheses show the p values. Variables and sources are the same as in Table 3 and the authors’ calculations. In FMOLS, lags 2.00 are used, and the linear and quadratic eqtrend effects are controlled (2). In DOLS, AR lag is not used.

, to consider past effects. Finally, the fully modified and dynamic OLS methodologies were used to deal with the possible endogeneity problem, considering the variables’ past and future temporal effects.

Key Conclusions of the Empirical Model

The positive relationship between the GDPpc and energy consumption was robust at the 1% significance level in each sample and with any econometric methodology, as shown in Table 4. Furthermore, this relationship was not affected by the method of estimation even when an autoregressive factor was included in the model, which increases the explanatory power. Using the FMOLS and DOLS methodologies, it can be seen that as the percentage of the LGDPpc increased by 1%, energy consumption increased by 0.6 to 0.65%. This conclusion was consistent with the literature [5,9,13]. The effect was considered significant. As economic growth improves in a country, there is a greater demand for energy consumption in all aspects of human activity (transport, heating, health, etc.). Moreover, as corruption increases, so too does energy consumption. As mentioned above, corruption affects energy consumption in various ways [10,30,31]. In our model, the growth factors prevailed; an increase of 1% of corruption caused an increase of 0.20 to 0.40% of energy consumption.

Additionally, concerning the remaining explanatory variables, the increase in renewable energy consumption led to a rise in total energy consumption, as energy demand is still high worldwide. An increase of 1% of renewable energy use caused an increase of 0.35 to 0.50% of energy consumption [45]. Energy efficiency, improved with innovation, is expected to reduce energy consumption. However, increased energy efficiency will increase the demand for energy consumption. When energy consumption rises, it compensates for the partial or completely reduced energy consumption caused by improved energy efficiency [81]. The final energy consumption will increase, and the rebound effect of energy efficiency will occur. For example, due to the use of energy-saving technology, the fuel consumption of transport is reduced, and people are more willing to frequently transport. In our model, an increase in innovation (RDEXP) causes an increase in energy consumption [25].

An essential factor affecting energy consumption is the temperature of the winter months (TEMPWIN). There must be high energy consumption in countries where this temperature is low. This happens in the north countries, which have very high energy consumption levels due to heating needs (Iceland, Finland, Sweden, Canada, etc.). Indeed, reducing this temperature by an average of one degree Celsius causes an increase in energy consumption of almost 2% [5]. A similar correlation is possible and vice versa in the summer months, but this has not been confirmed. We also investigated the factors that cause economic growth. We found that the increase in total investment causes a reduction in energy consumption due to greater efficiency and the shift of investment to renewable sources. Population growth and freedom of trade cause a statistically significant increase in energy consumption due to increased consumer power.

We studied the relationship between financial development (FINDEV) and energy consumption. Knowing from the literature that there are two opposite tendencies (increase due to greater access to energy products and decrease due to investments in renewable energy sources), we investigated the effect to declare the causality in the next section. We found that an increase in private sector financing of 1% of GDP causes an increase in energy consumption of 0.35 to 0.45%, which means that greater accessibility to high energy sources prevails for the research period [59]. We must point out that the periods after an economic crisis are usually aggressive, and high-efficiency solutions are preferred over other, more qualitative, and long-term ones. In contrast, Chiu and Lee [60] investigated the relationship over a long period (1984–2015), and found a negative relationship between the variables.

In addition, we explored whether energy losses in its production and distribution affect its total consumption, and we found that the effect is negative but not statistically significant [76]. The negative sign is justified because the relatively rigid supply cannot directly compensate for the losses. Finally, we investigated whether political stability (PS) affects the energy consumption of our model, and we did not find a statistically significant result in the whole sample for the period under investigation.

4.3. Model of Factors That Affect Electricity Consumption

We investigated the effect of independent variables on electricity consumption to find: (1) the coherence of the model and (2) the differences between electricity and total energy consumption. We applied the same econometric methodologies to the model for robustness, with the electricity consumption (kWh per capita) as a dependent variable.

Table 5. Electricity consumption and its determinants: results with the PCSE, FMOLS, and DOLS estimation methods.

Dependent Variable	Electricity Consumption (LENC) 981 Observations, 109 Countries (Total)
Model	PCSE	PCSE (ar1)	FMOLS	DOLS
ENRGLOS	−0.030186 *** (0.000)	−0.016665 *** (0.000)	−0.039364 *** (0.000)	−0.031739 *** (0.000)
LGDP	0.954626 *** (0.000)	0.982630 *** (0.000)	0.807013 *** (0.000)	0.941005 *** (0.000)
RENW	−0.000013 (0.968)	−0.001893 ** (0.048)	−0.001005 (0.564)	−0.000011 (0.997)
INVT	−0.008499 *** (0.000)	−0.003909 * (0.074)	−0.022739 *** (0.000)	−0.006854 (0.347)
POP	0.015929 (0.267)	0.006193 (0.711)	0.019459 (0.561)	0.022501 (0.668)
LCPI	0.234156 *** (0.000)	0.043384 (0.253)	0.367298 *** (0.001)	0.255695 ** (0.026)
TEMPWIN	−0.022183 *** (0.000)	−0.016443 *** (0.000)	−0.028172 *** (0.000)	−0.022586 *** (0.000)
RDEXP	0.025375 *** (0.004)	0.040826 (0.136)	0.015270 (0.750)	0.022331 (0.749)
FINDEV	0.002260 *** (0.000)	0.001479*** (0.003)	0.001020 (0.328)	0.002566 * (0.091)
PS	0.066165 *** (0.000)	0.016279 (0.503)	0.179747 *** (0.000)	0.078145 (0.300)
$R^2$	0.8816	0.9739	0.4280	0.8827

*, **, and *** mean the significance level of 10%, 5%, and 1%, respectively, the numbers in parentheses show the p values. Variables and sources are the same as in Table 3 and the authors’ calculations. In FMOLS, lags 2.00 are used, and the linear and quadratic eqtrend effects are controlled (2). In DOLS, AR lag is not used.

shows the results of the regressions, the coefficient, and the significance level.

Based on the models, the following conclusions were drawn:

(1)

The regression model of electricity consumption from its determinants is more coherent and has greater interpretive power than that of total energy consumption ($R^2$ higher and greater agreement of the estimators at 45% instead of 30%).

(2)

The impact of economic growth is more substantial on electricity consumption than on total energy consumption (0.8% instead of 0.6%) [10]. One possible explanation is that the electricity demand is higher than other forms of energy in developing countries. Economic growth comes from developing countries that initially need efficient forms of energy.

(3)

Innovation does not significantly affect electricity consumption concerning overall energy consumption. It seems that innovative activities aim to substitute other forms of energy to a greater extent than in electricity, which remains unaffected. At the same time, we found that population growth does not have a statistically significant effect on electricity consumption.

(4)

The financial development affects the electricity demand to a lesser extent than the energy consumption (0.2% instead of 0.36%). The increase in private sector financing is directed towards other forms of energy, such as the co-financed renewable sources in many countries, especially the developed ones. Finally, we found that energy production and distribution losses significantly reduce electricity consumption.

4.4. Robustness Checks

4.4.1. Econometric Methodologies and Endogeneity

Table 4 presents the models through various econometric methods, giving similar estimators with stable signs to basic variables (columns 1, 2, 5, and 6). We used dynamic error correction methods and, due to endogeneity issues, the probability of omitted variables, and the identification of cointegration and cross-sectional dependence issues, the FMOLS and DOLS methodologies were also applied.

4.4.2. Different Variable Models

In columns 3, 4, and 5, we used the gradual completion of our final model. First, we used our four essential variables (column 3). We added the state intervention with the growth and financial development variables through INVT, POP, PS, and FINDEV (column 4) and introduced corruption, innovation, trade openness, and winter temperature (column 5). We found stability in the standard variables and a proportional increase in the completeness of the models, as $R^2$ gradually increased from 0.23 to 0.29. As the independent variables are introduced step by step, the model acquires more interpretive power.

4.4.3. Alternative Variables for Corruption and Innovation

Next, we used different variables to assess corruption and innovation and found that the directions of impact and the size of the coefficient did not change significantly. We found something similar in the previous analyses, when replacing energy consumption with electricity consumption.

Table 6. Energy consumption and its determinants: alternative variables for corruption and innovation with the FMOLS estimation methods.

Dependent Variable	Energy Consumption (LENC) 981 Observations, 109 Countries (Total)
Model	MODEL 1	MODEL 2	MODEL 3	MODEL 4
ENRGLOS	−0.000029 (0.988)	−0.004874 ** (0.043)	−0.004344 * (0.061)	−0.000179 (0.968)
LGDP	0.662581 *** (0.000)	0.617629 *** (0.000)	0.679478 *** (0.000)	0.648066 *** (0.000)
RENW	0.003901 *** (0.000)	0.003267 *** (0.001)	0.004404 *** (0.000)	0.003959 ** (0.034)
PS	0.038866 * (0.095)	0.045026 (0.120)	0.016864 (0.536)	0.067593 (0.215)
FINDEV	0.004013 *** (0.000)	0.002741 *** (0.000)	0.003464 *** (0.000)	0.003743 *** (0.001)
INVT	−0.018238 *** (0.000)	−0.017024 *** (0.000)	−0.019815 *** (0.000)	−0.016801 *** (0.001)
POP	0.215477 *** (0.000)	0.214224 *** (0.000)	0.209344 *** (0.000)	0.200961 *** (0.000)
LCPI			0.205002 *** (0.000)	0.379376 *** (0.001)
LCCI	0.288286 *** (0.000)
ICRG		0.118990 *** (0.000)
TROP	0.000567 ** (0.026)	0.000346 (0.272)	0.000493 * (0.094)	0.000311 (0.595)
TEMPWIN	−0.020106 *** (0.000)	−0.022044 *** (0.000)	−0.021898 *** (0.000)	−0.027155 *** (0.000)
RDEXP	0.171698 *** (0.000)	0.180716 *** (0.000)
LGII			0.118392 * (0.056)
PAT				0.00808 ** (0.043)
$R^2$	0.2771	0.3424	0.3368	0.3201

*, **, and *** mean the significance level of 10%, 5%, and 1%, respectively, the numbers in parentheses show the p values. Variables and sources are the same as in Table 3 and authors’ calculations. In FMOLS, lags 2.00 are used, and the linear and quadratic eqtrend effects are controlled (2).

shows the regression results using LCCI and ICRG index for corruption (replacing LCPI) and LGII and PAT index for innovation (replacing RDEXP).

4.4.4. Developing and Developed Countries

We split the sample into developed (53) and developing (56) countries to test the model’s strength and its correlations. For this separation, first, the definition of the World Bank was used, where the criterion was per capita income. The separation limit for GDPpc was USD 15,000, splitting the sample down the middle. The separation aimed to control the consistency of the sample to investigate whether our variables affected developing countries differently than developed countries.

Table 7. Distinguishing developing and developed countries. FMOLS estimation method.

	981 Observations, 109 Countries (Total)
Dep. Var.	Energy Consumption (LENC) 477 Observations, 53 Countries (Developing)		Energy Consumption (LENC) 504 Observations, 56 Countries (Developed)
Model	MODEL 1	MODEL 2	MODEL 3	MODEL 4
ENRGLOS	0.009296 * (0.060)	0.004928 *** (0.000)	−0.039956 *** (0.002)	−0.036927 *** (0.000)
LGDP	0.473276 *** (0.000)	0.485043 *** (0.000)	1.024971 *** (0.000)	1.339227 *** (0.000)
RENW	0.002307 (0.401)	0.006551 *** (0.004)	0.006825 ** (0.023)	0.008607 *** (0.000)
INVT	−0.001384 (0.819)	−0.018561 *** (0.000)	0.004654 (0.654)	0.031208 *** (0.000)
POP	−0.106804 * (0.057)	−0.065599 *** (0.000)	0.126327 ** (0.012)	0.116876 *** (0.000)
LCPI	0.879746 ** (0.031)	1.165522 *** (0.000)	0.238688 * (0.056)	0.396371 *** (0.000)
TROP	0.004065 * (0.051)	0.006683 *** (0.000)	−0.001665 ** (0.014)	−0.000738 * (0.060)
TEMPWIN		−0.007936 *** (0.000)		−0.018062 *** (0.000)
RDEXP		0.678643 *** (0.000)		0.052950 * (0.078)
FINDEV		0.000667 (0.301)		0.000040 (0.950)
PS		−0.005801 (0.742)		−0.041014 (0.408)
$R^2$	0.1016	0.1183	0.1170	0.0724

*, **, and *** mean the significance level of 10%, 5%, and 1%, respectively, the numbers in parentheses show the p values. Variables and sources are the same as in Table 3 and the authors’ calculations. In FMOLS, lags 2.00 are used, and the linear and quadratic eqtrend effects are controlled (2).

shows the estimators of these models.

From the results of the estimators, we observed the consistency of the model and its strength. The signs of many variables remained in the same direction. However, there were also significant differences between the two subsamples.

GDPpc has a more significant impact on energy consumption in developed countries than in developing ones (1% increase instead of 0.5% when GDPpc increased by 1%). The most probable reason is that the structures are compact in developed countries and have an integration phase. Every form of development has a direct impact and comes from the private sector. In addition, we note that the increase in investment has a positive effect on the overall energy consumption in developed countries and a negative one in developing countries. This phenomenon deserves further investigation, as it should also have a negative sign in developed countries due to investments in renewable energy sources. One possible reason could be the research period (2010–2018). During this period developed countries, recovering from the global financial crisis, utilized all productive potential opportunities, including experiencing migration from neighboring developing countries.

Population growth increases energy consumption in developed countries while decreasing it in developing ones (−6.5% instead of 11%). The main reason is that there is a strong possibility of increasing energy consumption at constant prices in the developed ones, whilst this is only a doubtful possibility in developing countries with supply restrictions.

At the same time, rising corruption has less of an impact on energy consumption in developed countries than in developing ones (1% instead of 0.25%) [84]. As the positive sign prevails, the direct effect of corruption outweighs the indirect negative effect. In this case, the developed countries have dealt better with the problem of corruption; therefore, the corresponding effect is minor.

Finally, trade openness has a negative impact on energy consumption in developed and a positive impact in developing countries. The main reason for this is that more renewable energy sources are introduced in developed countries through foreign trade. In contrast, in developing countries, more importance is given to other basic goods for the state of living and development. Moreover, energy losses cause a substantial reduction in energy consumption in developed countries instead of developing ones, in which there is an increase (−3.5% instead of 0.5%). The losses in the former can possibly be replenished from renewable sources, while in the latter from the normal ones.

4.5. Results of Granger Causality Test

Dumitrescu and Hurlin [26] proposed an extension of the Engle and Granger [85] model, arguing that the proof of the occurrence of an integrated vector between variables might be followed by a causal relationship between them in one or both directions in the panel data. We explored the causal relationship between the variables of energy consumption and economic growth, corruption, innovation, and financial investment, so that the presentation of these relationships would be more informative. There was no test performed to determine the time lags because we applied the test for only one period (lag (K) = 1), but we transformed the data for nine time periods (T > 5 + 3K).

Table 8. Results of Granger Causality tests as Dumitrescu and Hurlin [26] performed.

α/α	Null Hypothesis (${\mathit{H}}_0$)	W-Bar Statistics	Z-Bar Tilde Statistics	p-Value
1	ENRGLOS does not Granger-cause LENC	3.3477	3.7229 ***	0.0002
2	RENW does not Granger-cause LENC	2.7855	1.4217	0.2706
3	LENC does not Granger-cause RENW	3.4602	1.3327	0.1643
4	TEMPWIN does not Granger-cause LENC	2.7821	2.4704 **	0.0135
5	LGDP does not Granger-cause LENC	3.2585	3.5254 ***	0.0004
6	LENC does not Granger-cause LGDP	3.9958	5.1585 ***	0.0000
7	LCPI does not Granger-cause LENC	1.7065	0.0882	0.9294
8	LCCI does not Granger-cause LENC	2.2989	1.5232	0.1277
9	INVT does not Granger-cause LENC	3.3726	3.7781 ***	0.0002
10	POP does not Granger-cause LENC	4.0583	5.2969 ***	0.0000
11	RDEXP does not Granger-cause LENC	2.6005	2.0683 **	0.0386
12	FINDEV does not Granger-cause LENC	4.5321	6.3462 ***	0.0000
13	LENC does not Granger-cause FINDEV	2.4316	1.6941 *	0.0903
14	PS does not Granger-cause LENC	4.6527	6.6131 ***	0.0000

*, **, and *** mean the significance level of 10%, 5%, and 1%, respectively, the numbers in parentheses show the p values. Variables and sources are the same as in Table 3 and the authors’ calculations. We used the test with lag = 1.

shows the corresponding values for the variables of interest, with the null hypothesis presented, the statistic, and their significance:

As shown in row 1, there is a unidirectional causal relationship between the energy losses and the energy consumption, which was expected. When the losses in energy grow, the anelastic energy supply cannot easily replace the losses in short periods.

Lines 2 and 3 show no causal relationship between renewable energy sources and total energy consumption. It transpired that the main reason for this is the different directions of the regression in developing and developed countries. Line 4 shows the causal relationship of the change in winter temperatures to the change in energy consumption, fully confirming the empirical analysis. There was a bidirectional causal relationship between economic growth and energy consumption [9,41], as economic growth affected energy consumption, and the latter is an essential determinant for growth (rows 5–6). Additionally, we found that corruption has no causal relationship with energy use (rows 7–8).

It also seemed that the relationship between investment and population growth with energy consumption was causal (rows 9, 10). These conclusions align with the empirical results, because there was a statistically significant coefficient in all models and any econometric methodology for the reasons analyzed above. In addition, line 11 shows the causal relationship between innovation and total energy consumption. This is often addressed to the production and distribution methods of energy by households and businesses.

Similarly, financial development has a causal bidirectional relationship with energy consumption. This relationship is more robust when directed from the first variable to the second than vice versa (rows 12, 13). This was also expected, and confirmed the empirical results and the literature [48,49,50], which state that significant financial improvement leads to increased energy consumption under any circumstances. Even if it is aimed at renewable sources, substantial funding leads to an overall increase in energy consumption, generally boosting growth. When energy consumption increases, one of the results is higher funding requirements. Finally, row 14 shows a unidirectional causal relationship between political stability and energy consumption, which is evident in the empirical model but has not been strongly demonstrated. The main reason is that the political stability variable is also correlated with other macro variables in the model, within the acceptance limits, but which affects the overall correlation.

5. Conclusions and Policy Implications

The incomplete and contradictory results found in the studies on the relationship between energy consumption and its causes call for the need to open new perspectives for exploring the relationship between economic growth, energy consumption, innovation, and financial expansion. We have contributed to this literature by introducing a comprehensive model of institutional quality, which corresponds to the level of corruption, climate as proxied by temperature fluctuations, and financial expansion expressed through financial development. The present study could, therefore, be considered as innovative; although it uses known econometric methods, it introduces a set of several variables into a comprehensive and holistic model of interpretation and prediction for the benefit of the current researchers in the energy environment and development literature.

We use the panel data from 109 countries from 2010 to 2018 and classify the data into two income groups based on the income level and the World Bank definitions. It was found that the causal relationship in each group is relatively different. To have uncorrelated variables, we selected the basic ones and one out of each category of the causes. For example, we chose the winter temperature and not the summer temperature for a particular time as well as the demand and energy supply and not the prices of products, and, finally, the population growth and not urbanity. Using data from all countries as a whole, we discovered that there is a bidirectional positive relationship between economic growth and energy consumption.

The estimated results were robust to different econometric specifications. Each variable affected energy consumption directly or indirectly and impacted the economy. For instance, an increase in the GDPpc increases energy consumption due to the greater availability of resources in households and businesses. An increase in corruption causes an increase in energy and electricity consumption because the increment effect of reducing energy policy stringency outweighs the decreasing impact on per capita output while reducing energy consumption. This effect was more pronounced in developing countries where the bureaucracy is a major restraining problem of the economy. In our research, in all models and sub-models, it was found that the increasing effect of corruption on energy consumption and all the corruption variables (CPI, CCI, ICRG) were of great importance. This critical finding reduces the existing doubts of the empirical literature on the issue [5,31]. In addition, improving innovation, investments, financial aid, and trade freedom leads to increased energy consumption and economic growth. The shift to renewable energy sources and the reduction of losses in the production and distribution of energy lead to the same result.

Therefore, a necessary process that needs to be implemented is the reduction of corruption. Due to its increasing impact on energy consumption, especially in developing countries, its decline will result in a lower energy burden without compromising the economic growth in the long term. This can be achieved by reconstituting the institution with more transparency and strict anti-corruption policies. A genuine commitment to the political leadership in this challenging area would be required because corruption is a complex social phenomenon. In particular, the organizations that provide licenses for energy products and the companies that manage the production and distribution of energy must be controlled with strict rules and frameworks. The specific market must be transparent and its operating framework clear. At the same time, specialized investments in the energy sector would be required through development programs and the strengthening of innovative products in production, distribution, and energy use. This can be achieved through subsidies, tax exemptions, and the simplification of their framework.

This process of reforming the framework is not without consequences. On the one hand, transparency and control are required, while on the other hand, simplification is essential. At the same time, it seems that in addition to stringent rules, financial resources are also necessary. Otherwise, this might lead to a short-term stagnation in economic growth. Still, the long-term gains will be significant, as reducing corruption and increasing innovation and professional investment will generally improve workforce efficiency and transparency, thus attracting a healthy investment.

Furthermore, the findings reveal that the main determinants for energy consumption in developed countries are economic growth, investment, population growth, and winter temperature. In the developing ones, the determinants are trade openness, corruption, and innovation. These results show that different policy actions are needed in developing and developed countries. Therefore, in the developed nations, the maintenance of growth, investment, and the quality and quantity of the labor force must be given priority, while in the developing ones, the treatment of the bureaucracy, the improvement of infrastructure, and the bettering of the regulations that hinder free trade are required.

In addition, a bidirectional relationship between GDPpc and energy consumption is found to exist. Thus, any effort to improve economic growth leads to increased energy consumption and climate change, thus contributing to energy demands. This means that a development policy, as well as a restrictive energy policy, are required. As such, an energy conservation policy might be feasible without causing harm to the GDP. The above empirical analyses show that climate plays a more critical role in energy consumption than the other factors. For example, when winter temperatures are reduced by 1 °C, this leads to an increase in energy consumption by at least 2%. Therefore, efforts to improve institutional quality by providing financial assistance or distribution channels and reducing energy production and distribution losses could be relatively ineffective for improving energy efficiency. The results also indicate some other important policy implications. In economies where the above situation applies, policymaking should focus on decoupling energy from economic growth by developing newer technologies and improving human capital.

A solution to this dual problem is to boost growth by producing energy that does not harm the climate. This essentially implies a gradual transition to renewable energy sources, which would promote economic growth without climate change. Therefore, governments need to ensure an adequate supply of available energy to economies and take appropriate measures to create effective financial development policies and institutional frameworks for significant technological change. Large investments should be made in the renewable energy sources, and also simplifications in their licensing framework, in order to increase the use of solar panels and other renewable products from companies and households. At the same time, educational and public information programs are required, in order to persuade people to use energy more carefully and to benefit from renewable sources. Additionally, they must provide sufficient investment in research and development to stimulate human capital to drive the required knowledge in economies to ensure efficiency in the energy and financial sectors. The results also indicate a 1% increase in financial development, causing an increase in energy consumption by 0.3%. Therefore, the banking system and monetary authorities must encourage investments in energy-efficient technologies and provide the optimal use of the financial instruments and technologies available in the system to stimulate economic growth and energy efficiency. There must be specialized funding for innovative renewable energy programs so that the actions of the monetary authorities are combined with those of the government.

In conclusion, policymakers need to carefully implement economic and developmental policies and promote more efforts towards environmentally friendly energy usage and thus reduce the detrimental effects of the economic developmental strategies.

The present research was limited to the period between 2010 and 2018, regarding the data after the recent financial crisis. Further research could include a more extended period, so that conclusions on causation could be more secure. At the same time, a survey could split the sample into countries with restrictive programs as well as those without such programs, so that differences can be scrutinized accordingly.