Introduction

One of the grand challenges policymakers around the world face is to address climate change while meeting the demand for energy consumption. The worldwide energy demand and pattern of total energy consumption have radically changed because of increases in industrial production and the world’s population. The global demand for energy is projected to increase by 30% by the year 2040 (IEA 2017). Because of this increase in the demand for energy, measures to increase sustainable and cost-efficient clean energy sources must be made to control the issues of environmental degradation and attain sustainable economic growth across the globe. The challenge of achieving this objective requires action plans and supportive policies to develop efficient and effective renewable energy resources (Han et al. 2017; Wiser et al. 2016) that can contribute to a sustainable energy, economic, environmental, and societal structure (Kuriqi et al. 2017). In a comprehensive empirical study of the effects of various factors on renewable energy demand, Aguirre and Ibikunle (2014) illustrated that carbon dioxide (CO2) emissions, energy consumption, and gross domestic product (GDP) per capita can influence the use of renewable energy in the effort to implement the Kyoto protocol. A spatial spillover effect on production of renewable energy was recorded in the countries of the EU (Shahnazi and Shabani 2020). In addition, the EU countries have endorsed targets to attain a 20% share in consumption of renewable energy by 2020 and to increase that share up to 32% by 2030.Footnote 1 The policy related to renewable energy sources in the EU has demonstrated the importance of renewable energy consumption since 2000.

A sustainable and renewable energy system is a significant improvement over a conventional energy system because it uses resources that can be replenished (Apergis and Danuletiu 2014). The use of renewable energy sources is a potential solution to the climate change crisis and issues of energy security around the world (Elliot 2008). Paramati et al. (2017a, b) empirical study indicated that G20 countries substantially reduce CO2 emissions through the consumption of various sources of renewable energy. Thus, in developed countries, CO2 emission levels can be the primary driving force in demanding consumption of clean energy (Omri and Nguyen 2014). More recently, Gozgor et al. (2020) found that CO2 emissions per capita have a positively impact on the consumption of clean energy in thirty OECD countries.

A sustainability strategy is considered long-term when it includes investment in green technology to increase future economic growth (Hart and Dowell 2011). Increasing restrictions on the use of fossil fuels help to overcome climate change and motivate sustainable development through the transition from the conventional energy sector to the clean energy sector (Gallo et al. 2016). The Bloomberg New Energy Finance BNEF (2020) reported that, in 2019, the worldwide investment in clean energy sources was $282.2 billion, a 1% increase from 2018.

The growing concern about sustainable economic development has generated research interest in examining the importance of renewable energy consumption. For instance, Stiglitz (2002) stated that sustainable development is attained through various aspects but renewable energy is the fundamental synergy factor. Kaygusuz (2007) also indicated that renewable energy promotes a continuous process of modernization in the energy sector and supports the goals of sustainable economic development in various countries. Studies have documented that per capita GDP growth is one of the main determinants of the positive affect of clean energy consumption (Dogan and Ozturk 2017; Kahia et al. 2016; Sadorsky 2009a, 2009b). In a study of the casual link between renewable energy use and economic growth, Saad and Taleb (2018) reported unidirectional causality in the short run and bidirectional causality in the long run between economic growth and renewable energy use in twelve EU countries.

Research has also inspected the relationships among institutional quality, green energy consumption, and environmental quality. For instance, García-Álvarez et al. (2018) examined the EU’s renewable energy policies based on three aims—sustainability, competitiveness, and security—for the period from 2000 to 2014 and found that elements of governmental policy like quota, contract duration, and tariff size have positive influences on the production of clean technologies. This stream of research has suggested that green energy consumption and carbon emissions may depend on certain institutional and economic conditions, including the rule of laws, corruption, bureaucratic quality, state disclaimers of contracts, and risk of expropriation. For example, Callway (2013) study of political and economic issues identified higher frontier costs, credit repayment terms, variations in subsidies, and the taxation system as key obstacles in the investment and development of renewable energy and its consumption. Recently, Uzar (2020) investigated a panel of 38 countries for the period from 1990 to 2015 and reported a positive effect of institutional quality and CO2 emissions on renewable energy consumption, but a negative impact of GDP growth on renewable energy. Market incentives like research and development grants, tax incentives, lower financing rates, and lower insurance premiums ease the adoption of clean energy technology (Aragón-Correa and Sharma 2003).

Shahbaz et al. (2018a, b, c) found that globalization promotes financial development, trade openness, and economic development across the globe. They argued that foreign trade, industrialization, increasing investment, and urbanization to attain a high level of economic development cause pollution and overall environmental degradation. Çoban and Topcu (2013) reported that growth in financial development has a significant positive influence on energy use among the older member countries of the EU. Finally, Gozgor et al. (2020) empirical investigation of thirty OECD countries found that the economic factors of globalization contribute to enhancing the use of renewable energy. The EU countries lead the world in the use and research and development of renewable energy (Halicioglu and Ketenci 2018). Therefore, this study analyzes EU-28 countries’ contemporary challenge of increasing their clean energy consumption in terms of CO2 emissions, clean energy technology, GDP growth, institutional quality, and globalization.

Since few empirical studies have tested the impact of environment degradation, rapid economic growth, and globalization on clean energy consumption, our study contributes to the literature by investigating the effects of clean energy technology and institutional quality on clean energy consumption in the EU-28 countries. This study differs from others, first, in that it includes EU-28 countries’ public and private sector investment in renewable technologies integrated into the system, new technologies and services for consumers, the resilience and security of the energy system, new materials and technologies for buildings, energy efficiency for industry, a competitive global battery sector, decarbonization technologies, renewable fuels, nuclear safety, and others. Thus, this study provides a broader and better description of the role of clean energy technology in the use of clean energy. Second, this study is the first to explore the effects of clean energy technology on clean energy consumption in the EU countries. Third, this research paper employs two long-run analysis methodologies, the continuous updating-fully modifying (CUP-FM) technique, and the continuous updating-bias correcting (CUP-BC) technique. We validate our study findings by applying Driscoll and Kraay’s (DK) standard error technique. Therefore, this study provides a broad and reliable picture of the factors involved in clean energy consumption.

The rest of this study proceeds as follows: the “Literature review” section describes the nexus among all study variables. The “Materials and methodology” section discusses the data sources and describes the variables and data analysis techniques. The “Results and discussion” section reveals the results of the data analysis and the various analysis techniques. Finally, the conclusion concludes the study and suggests its policy implications.

Literature review

Given the important role of clean energy consumption in sustainable economic growth, clean energy meets many countries’ need for energy and is important in mitigating the issue of CO2 emissions. Tang and Tan (2015) reported a casual association between energy consumption and CO2 emissions in Vietnam using annual data for the period from 1976 to 2009. Several empirical studies have investigated the role of the relationship between clean energy technology and clean energy consumption on environmental degradation. The findings of the limited literature that has explained the association between these variables have been ambiguous and contrasting. Therefore, Balcilar et al. (2018) argued that new studies are required to validate and explicate the existing literature and address these current contrasting findings. Various studies have indicated that technological innovation is necessary if a country is to face ecological challenges and mitigate environmental degradation (Alvarez-Herranz et al. 2017; Andreoni and Levinson 2001; Lorente and Álvarez-Herranz 2016). The nexus between the use of energy and economic growth can be tested through four categories of hypotheses: conservation, feedback, growth, and neutrality (Apergis and Payne 2012). Many studies have supported the feedback hypothesis by reporting bidirectional causality between renewable energy use and economic growth in both the short run and the long run (Apergis and Payne 2011; Kahia et al. 2016; Sebri and Ben-Salha 2014).

Khoshnevis Yazdi and Shakouri (2017b) showed unidirectional causality between renewable energy consumption and economic growth, and Dogan and Ozturk (2017) argued for unidirectional causality between economic growth and consumption of renewable energy in the short run and in the long run, finding support for the feedback hypothesis through findings of bidirectional causality. Institutional voids like underdeveloped infrastructure and inadequate rules, regulations, and law enforcement generate barriers and uncertainty in the business environment (Mair et al. 2012). Studies have shown the negative impact of technological innovation when institutions like these are absent or weak (Michailova et al. 2013; Zhu et al. 2012). Adoption of clean energy technology is eased by market incentives like research and development grants, tax incentives, low bank financing rates, and low insurance premiums (Aragón-Correa and Sharma 2003). Many economic and market incentives are also connected to policy instruments; for instance, credit policies for emission trading influence investment in low-carbon and green technologies (Wordsworth and Grubb 2003).

The literature has provided three main reasons for increased energy consumption that is due to globalization channels. First is the impact of scales and the argument that the positive correlation between globalization and energy consumption is due to increased economic activities (Cole 2006). Second is the impact of technologies, which suggests that globalization, rather than lowering the level of economic activity, works as a motivation to import new technologies that reduce energy consumption (Dollar and Kraay 2004). Third is the impact of consumption, as globalization reduces the energy consumption that is due to increased economic activities (Stern 2007).

Globalization is measured by proxies like trade openness, imports, exports, and trade liberalization, proxies that have also been used to assess the connection between energy consumption and globalization (Shahbaz et al. 2016). Normally, globalization expands with increased energy consumption because of the high level of economic growth. This view is observed in the literature, although other studies report a reverse influence of globalization on consumption of energy. For instance, Shahbaz et al. (2018a, 2018b, 2018c) used a panel data set of 25 economies to find a positive link between energy consumption and globalization in 12 countries, but a negative link in the UK and the USA. The literature review of studies that have addressed the nexus between clean energy consumption, CO2 emissions, clean energy technology, GDP growth, institutional quality, and globalization is summarized in Table 1.

Table 1 Summarized literature review
Full size table

Materials and methodology

Methodological framework

The primary objective of this study is to test the long-run associations among clean energy consumption, CO2 emissions, clean energy technology, GDP growth, institutional quality, and globalization. The paper conducts empirical analyses of panel cointegration, tests elasticities for long-run associations, and tests for non-causality of heterogeneity to identify the direction of causal relationships among the study variables. Equation 1 is based on a benchmark model:

$$ CEC=f\left(C{O}_2+ CET+ GDP+ QOI+ GOB\right) $$
(1)

where CEC, CO2, CET, GDP, QOI, and GOB refer to clean energy consumption, CO2 emissions, clean energy technology, GDP, institutional quality, and globalization, respectively.

We have used the log-linear by employing natural logs for all study variables, rather than using the simple linear form of the model. Studies like Shahbaz et al. (2012) have argued that the empirical results estimated through log-linear are more reliable and consistent than simple linear. The empirical models’ log-linear provides direct estimations of elasticities because it works as the coefficients of the study’s explanatory variables of the study. The log-linear form of clean energy consumption function is presented in Eq. 2:

$$ \ln CE{C}_{it}={\alpha}_0+{\alpha}_1\ln C{O}_{2 it}+{\alpha}_2\ln CE{T}_{it}+{\alpha}_3\ln GD{P}_{it}+{\alpha}_4\ln QO{I}_{it}+{\alpha}_5\ln GO{B}_{it}+{\mu}_{it} $$
(2)

where ln is the natural log form of the variables in Eq. (2); α0 is the slope intercept; α1, α2, α3, α4, and α5 are the coefficient estimates of CO2 emissions, clean energy technology, GDP growth, institutional quality, and globalization, respectively. Error term is represented by μ, which is normally distributed. Subscript i (i=1,......., N) is the country, and subscript t (t=1,.......,T) is the time period.

Cross-sectional dependence (CD) and panel unit root tests

Empirical studies have recommended the characteristics and properties of variables in a time series data set; the main property is stationarity vs. non-stationarity. First- and second-generation unit root tests are used to examine this property, but the selection of a particular test is based on the assumption of cross-sectional independence. In general, the variables of panel data of many countries are linked because of regionally and global associations. If researchers fail to measure the assumption of cross-sectional independence, the chances of misleading estimated results are high (Phillips and Sul 2003). Therefore, we investigate cross-sectional dependence using Pesaran (2004) test of cross-sectional dependence. The test is performed with the following Eq. 3:

$$ CD=\sqrt{\frac{2T}{N\left(N-1\right)}}\left(\sum \limits_{i=1}^{N-1}\sum \limits_{k=i+1}^N{\rho}_{ik}\right) $$
(3)

where T and N are the time period and the sample size, respectively. Correlations between the error terms of different cross-sections of a country i and k are indicated by ρik.

After collecting evidence of cross-sectional dependence among study variables through these tests, this study used second-generation panel unit root tests to examine the residual stationarity in the presence of cross-sectional dependence because first-generation tests can give indecisive estimations when there is cross-sectional dependence (Dogan and Seker 2016). The panel unit root test in this study is performed using cross-sectionally augmented IPS (CIPS) and cross-sectionally augmented ADF (CADF). The unit roots tests of CIPS and CADF are applied using Eqs. 4 and 5, as Pesaran (2007) suggested:

$$ \varDelta {Y}_{it}={\gamma}_{it}+{\chi}_i{Y}_{i,t-1}+{\lambda}_iT+{\sum}_{k=1}^n{\pi}_{ik}\varDelta {Y}_{i,t-k}+{\mu}_{it} $$
(4)

where Δ is a difference operator; Iit are variables used in the empirical analysis; and T, γ, and μit are the time trend, the individual intercept, and the error term, respectively.

Next, CADF test is investigated using the standard augmented Dickey-Fuller (ADF), which add the averages of lagged levels’ cross-sections (\( {\overline{X}}_{t-1} \)) and uses the first difference values of an individual series. Thus, the CADF test uses Eq. 5:

$$ \varDelta {X}_{it}={\alpha}_i+{\beta}_i{X}_{i,t-1}+{\delta}_i{\overline{X}}_{t-1}+\lambda \varDelta {\overline{X}}_t+{\mu}_{it} $$
(5)

where \( {\overline{X}}_t \) is the average values of all available N observations in the sample at time period t. This equation includes a proxy to measure unobserved effects through common factors.

Panel cointegration test

If the levels of the study’s variables have no stationarity, then a cointegration test of the variables is used to ensure the economic and statistical accuracy of the coefficient estimations through long-run analysis. To determine whether cointegration exists between clean energy consumption, CO2 emissions, clean energy technology, GDP, institutional quality, and globalization, this study uses a bootstrap test for cointegration as provided by Westerlund and Edgerton (2007). Equation 6 is used for the bootstrap test:

$$ {\displaystyle \begin{array}{c}{y}_{it}^{\ast }={\hat{\alpha}}_i+{x}_{it}^{\ast^{\prime }}{\hat{\beta}}_i+{z}_{it}^{\ast}\\ {} with\\ {}{x}_{it}^{\ast }=\sum \limits_{j=1}^t\varDelta {x}_{ij}^{\ast },\end{array}} $$
(6)

where \( {\hat{\alpha}}_i \) and \( {\hat{\beta}}_i \) are determined through the fully modified terms of αi and βi. The bootstrap test’s null hypothesis is that the variables of a panel data set are cointegrated. The small sample size also covers by this test and is suitable to allow all cross-sectional units’ dependency in cases of both between and within. Furthermore, problems like dependence of cross-sections and heterogeneity during the estimation procedure of cointegration between variables are also controlled through the bootstrap test.

Long-run elasticities

The long-run elasticities estimation between independent and dependent variables is done using two estimation techniques as proposed by Bai and Kao (2006) and Bai et al. (2009). Equation 7 is employed to determine two estimators that can overcome the issues of bias that result from dependence of cross-sections, serial correlation, and endogeneity:

$$ \left({\hat{\beta}}_{CUP},{\hat{F}}_{CUP}\right)=\mathrm{argmin}\frac{1}{n{T}^2}\sum \limits_{i=1}^n{\left({y}_i-{x}_i\beta \right)}^{\prime }{M}_F\left({y}_i-{x}_i\beta \right) $$
(7)

where repeated fully modified least squares (FM-OLS) are applied to measure the β coefficient, as FM-OLS uses previous stage residuals until full convergence occurs. The terms F and MF = IT − T−2FF', IT show a common factor, which is presumed by the dimensions of matrix T and error terms, respectively. Hence, F allocates the initial estimations and continues this process until all convergence is complete. The CUP-FM and the CUP-BC both continuously update until the convergence is complete (Bai et al. 2009). These two estimators provide consistent and unbiased results even in the case of exogenous regressors. Moreover, both estimators help to control issues I(1)/I(0) of mixed factors and establish robust outcomes. The FM-OLS procedure is followed by both estimators, so they provide consistent findings even in the absence of endogeneity (Bai et al. 2009).

Long-run elasticities results are also estimated by applying the DK standard error technique to investigate the effect of the study’s variables on clean energy consumption for a panel of EU-28 countries. Before the DK standard error technique can be employed, the product average among errors and independent variables must be calculated, and these calculated values used in estimating weighted heteroskedasticity and autocorrelation consistent (HAC) to determine standard errors. Doing so will help to deal with cross-sectional dependence (Jalil 2014).

The DK standard error technique is considered a preferred method, even cases of serial and spatial dependency or heteroscedasticity in the data set (Ozokcu 2017; Sarkodie and Strezov 2019). The technique allows all dimensions of large time and is flexible because it is a non-parametric approach. In addition, the DK technique works as a covariance estimator that handles missing values and it can be applied to either balanced as well or unbalanced panel data. The technique’s estimators are robust to general procedures of temporal and cross-sectional dependence. This study uses the DK standard error technique as a robustness test by using Eq. 8, a linear model equation expression of pooled ordinary least squares (OLS):

$$ {y}_{i,t}^{\prime }={x}_{i,t}^{\prime}\propto +{\mu}_{i,t},i=1,\dots, N,t=1,\dots, T $$
(8)

where \( {y}_{i,t}^{\prime } \) is the study’s dependent variable (clean energy consumption) and \( {x}_{i,t}^{\prime } \) is the independent variables (CO2 emissions, clean energy technology, GDP, institutional quality, and globalization).

Heterogeneous panel causality test

Econometric methods for measuring long-run elasticities estimate only the associations between dependent and independent variables, but policymakers also require short-run analysis to estimate the directions of causal relationships among study variables. Therefore, to determine the direction of casual associations between the dependent variable (clean energy consumption) and independent variables (CO2 emissions, clean energy technology, GDP, institutional quality, and globalization), this study uses an advanced procedure for a simple test of Granger causality that Dumitrescu and Hurlin (2012) suggested.

The heterogeneous issues and unbalanced panel properties of T < N and T > N can handle through flexible characteristics of Dumitrescu and Hurlin test. Moreover, this test incorporates the standard regression form of Granger causality in case of cross-sections, along with differences and average values of all coefficients by all units in the various cross-sections. Equation 9, a bivariate model equation, was used to apply the causality test:

$$ {y}_{i,t}={\alpha}_i+\sum \limits_{k=1}^k{\lambda_i}^{(k)}\ {y}_{i,t-k}+\sum \limits_{k=1}^k{\beta_i}^{(k)}\ {x}_{i,t-k}+{\varepsilon}_{i,t} $$
(9)

where αi is the slope intercept, λi and βi are coefficients of the slope, and k is the lags length in numbers.

Data and variables measure

A balanced panel data set was collected from the EU-28 countries: Austria, Belgium, Bulgaria, Croatia, Cyprus, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Poland, Portugal, Romania, Slovenia, Slovakia, Spain, Sweden, and the UK. The choice of time period of 1995 to 2017 was based on availability of annual data for the period. Measurements of the study variables and their data sources are given as follows.

Clean energy consumption (CEC)

Clean energy, or non-carbon energy, is produced through various renewable energy sources that do not produce CO2: hydropower, wind, solar, and geothermal. This study measures CEC as total renewable energy consumption in billion kilowatt hours (Kwh). The CEC data is collected from the U.S. Energy Information Administration (EIA 2019).

Carbon dioxide emissions (CO2)

The CO2 emissions in metric tons of a country are divided by the country’s total population to measure it in per capita values. This study uses CO2 emissions as a proxy for environmental degradation. The CO2 emissions data was gathered from the World Development Indicators (WDI 2019) database.

Clean energy technology (CET)

The CET is measured by combining the public and private investment in renewable energy research and development expenditures (in constant 2010 US dollars). The CET data was collected from the European Commission’s database.Footnote 2

Gross domestic product (GDP)

The GDP variable is measured in per capita values, dividing GDP figures (in constant 2010 US dollars) by the country’s total population. The WDI (2019) database was used to collect the GDP data.

Institutional quality (QOI)

The QOI is measured using countries’ economic freedom indices. The economic freedom index consists of the size of government, legal structure and property rights, ease of accessing sound money, trade policies and international trade, and the regulation of business, credit, and labor markets. QOI data was obtained from the Fraser Institute Index.Footnote 3

Globalization (GOB)

To measure the GOB, we used the globalization index, which is captured from KOF Swiss economic institute (Dreher 2006). The globalization KOF Index is a combination of three categories: economic, social, and political globalization. The data related to economic globalization consists of restrictions and actual flows; social globalization comprises the personal contacts, cultural immediacy, and the flow of information data; and political globalization includes factors like the country’s embassy relationships, international treaties, membership in international organizations, and participation level in missions of the UN Security Council. The KOF index is scaled between 0 and 100, where 0 indicates the country is not globalized, and 100 indicates it is completely globalized.

Results and discussions

Descriptive statistics results

The summary statistics of all variables included in this study such as clean energy consumption (total renewable energy consumption in billion kilowatt hours), CO2 emissions (CO2 emissions in metric tons per capita), clean energy technology (public and private investment in renewable energy research and development expenditures, constant 2010 US dollars), GDP growth (GDP per capita, constant 2010 US dollars), institutional quality (economic freedom index), and globalization (KOF index is scaled between 0 and 100) of each of the EU-28 countries is given in Table 1. The highest mean values for clean energy consumption (46.131), CO2 emissions (20.012), clean energy technology (5360.02), GDP growth (98098.55), institutional quality (8.141), and globalization (88.784) are in Sweden, Luxembourg, Germany, Luxembourg (again), the UK, and Belgium, respectively. The lowest mean values for clean energy consumption (0.391), CO2 emissions (3.406), clean energy technology (14.255), GDP growth (5769.521), institutional quality (6.551), and globalization (69.193) are in Estonia, Latvia, Greece, Bulgaria, Romania, and Latvia (again), respectively.

In addition, Luxembourg shows the highest variation from the mean in CO2 emissions and GDP growth, with values of 2.885 and 11292.58, respectively. Lithuania, Germany, Romania, and Croatia have the highest variation from the mean in clean energy consumption (17.227), clean energy technology (3296.68), institutional quality (1.257), and globalization (9.631), respectively. The EU countries with the lowest variation from the mean in clean energy consumption (0.187), CO2 emissions (0.301), clean energy technology (7.105), GDP growth (1295.441), institutional quality (0.079), and globalization (1.401) are Luxembourg, Latvia, Greece, Portugal, Germany, and Ireland, respectively (Table 2).

Table 2 Descriptive statistics
Full size table

Results of cross-sectional dependence and panel unit root tests

The analysis of panel data started with the cross-sectional dependence test Pesaran (2004) suggested. Table 3 presents the results of this test. After confirming the presence of cross-sectional dependence in the panel data, we examined stationarity in the panel data set using second-generation unit root tests. We employed the second-generation CIPS and CADF tests Pesaran (2007) proposed to control for the cross-sectional dependence. Table 4 shows results of the CIPS and CADF unit root tests. The first difference results of both tests indicate the presence of stationarity, so the panel data set has no unit roots at first difference.

Table 3 Analysis of cross-sectional dependence
Full size table
Table 4 Unit root tests of second generation
Full size table

LM bootstrap cointegration results

This study’s empirical analysis tests the cointegration among all of its variables using bootstrapping, as given by Westerlund and Edgerton (2007). The results of the technique are shown in Table 5. The null hypothesis cannot be rejected, so cointegration exists among the variables, supporting the long-run relationship among the variables clean energy consumption, CO2 emissions, clean energy technology, economic growth, institutional quality, and globalization.

Table 5 Results of testing LM bootstrap cointegration
Full size table

Results of long-run analysis

Several econometric methodologies can be used to measure the variables’ long-run elasticities. Our long-run analysis uses two of these: first, the CUP-FM and the CUP-BC methods (Bai et al. (2009) and then the DK standard errors technique. The results of the CUP-FM and CUP-BC techniques, provided in Table 6, show that CO2 emissions have a significant and positive impact on the consumption of clean energy, which supports the notion that countries are motivated to consume clean energy when their environmental pollution, measured as CO2 emissions, increases. These results echo the argument of Kusumadewi et al. (2017) in Thailand and Salim and Rafiq (2012) in Brazil, China, India, and Indonesia that renewable energy consumption mitigates the problem of increased environmental pollution.

Table 6 Panel long-run analysis
Full size table

These results show a positive and significant long-run relationship between clean energy consumption and clean energy technology. This finding is consistent with Dinda (2004) and Brock and Taylor (2005), who suggested that countries must use technologies that are based on environmentally friendly energy sources to control further environmental degradation. Further, GDP growth and energy consumption are generally considered to be among the main determinants of environmental pollution. Currently, policymakers focus on sustainable economic growth through the use of clean energy sources. This study reports that GDP growth has a positive and significant impact on clean energy consumption in EU countries, supporting Sadorsky (2009a) argument that economic growth plays a significant role in renewable energy consumption in eighteen emerging countries.

Wu and Broadstock (2015) argued that institutional quality has a positive and significant influence on renewable energy use, but this study reports a negative association between these two variables, perhaps because of the EU’s strict law enforcement, administration system, financial regulations, and taxation laws. The results of the CUP-FM and CUP-BC techniques show a positive and significant impact of globalization on clean energy consumption, which is consistent with Soytas et al. (2007) findings.

In the second part of long-run analysis, we apply the DK standard errors technique. The results of this regression test are given in Table 7. All reported coefficients in the DK standard errors technique show findings similar to those we found in using the CUP-FM and CUP-BC techniques. Figure 1 illustrates the key findings of our long-run analyses.

Table 7 Regression results of Driscoll-Kraay standard errors technique
Full size table
Fig. 1
figure 1

Key findings of long-run analysis

Full size image

Results of Dumitrescu and Hurlin causality test

This study examined the causality effects between all of its variables using a panel causality effect approach introduced by Dumitrescu and Hurlin (2012). The results of pair-wise panel causality, presented in Table 8, show that clean energy consumption and CO2 emissions have unidirectional causality, a finding that is consistent with studies by Ajmi et al. (2015) in G7 countries and Azlina et al. (2014) in Malaysia. The findings also show unidirectional causality between clean energy consumption and clean energy technology, which indicates that a continuous process in the development of clean energy technology is required to increase the use of clean energy. This finding supports findings in the studies of Lin and Zhu (2019) in China and Ganda (2019) in OECD countries, which argued that technological innovation in renewable energy is necessary to increase renewable energy consumption and control environmental pollution.

Table 8 Results of Dumitrescu and Hurlin heterogeneous panel causality test
Full size table

Clean energy consumption and GDP growth show unidirectional causality, which is in line with the findings of studies by Khoshnevis Yazdi and Shakouri (2017b) and Dogan and Ozturk (2017). Institutional quality and clean energy consumption show bidirectional causality, which matches the results of studies by Saidi et al. (2020) in MENA countries and Wu and Broadstock (2015) in 22 emerging economies. Globalization and clean energy consumption show unidirectional causality, which supports the findings of Apergis et al. (2010) in 19 developing and developed economies. Figure 2 shows the main causality effects.

Fig. 2
figure 2

Main results of causality effects

Full size image

Conclusion

Given the commitments and efforts of EU-28 countries to deal with environmental degradation issues by implementing sustainable strategies and increasing the consumption of clean energy, this study contributes to examinations of the effects of CO2 emissions, GDP growth, institutional quality, and globalization on clean energy consumption for the period from 1995 to 2017. To meet the main purpose of this study, we examined a panel data set using cross-sectional, panel unit root, and cointegration tests. We also employed CUP-FM and CUP-BC estimators, as suggested by Bai and Kao (2006) and Bai et al. (2009). Moreover, the study measures the validity and reliability of long-run coefficients using the DK standard error regression technique and determines the short-run causal relationships between variables by applying Dumitrescu and Hurlin (2012) heterogeneous panel causality test.

The findings of the long-run analyses show that CO2 emissions, clean energy technology, GDP growth, and globalization play a positive role in increasing clean energy consumption but a negative effect of institutional quality on clean energy consumption. The Granger causality test of the short-run causal connections between variables shows unidirectional causality between CO2 emissions, clean energy technology, GDP growth, and globalization with clean energy consumption. Institutional quality has a bidirectional relationship with clean energy consumption. In terms of environmentally friendly clean energy consumption, our study’s findings have useful implications, especially for EU-28 countries.

The study’s empirical findings have several policy implications. First, policymakers should understand the positive impact of CO2 emissions on environmental degradation and increase clean energy consumption. Second, the positive effects of clean energy technology show that EU countries are going in the right direction. However, they should maintain their investments in clean energy technologies if they are to achieve the commitment by the Council of the European Union (2014) that at least 27% of energy come from renewable sources by the end of 2030. In addition, by doing this, the EU can fulfil its’ commitment to achieving climate neutrality at the end of 2050 and to reaching target of dropping emissions of greenhouse gases by 55% at the end of 2030 in comparison to 1990 levels.Footnote 4

Third, future economic growth is based on continuous implementation of sustainable strategies through investing in green technologies. Therefore, the EU-28 countries’ governments should design policies that establish confidence among investors at both the domestic and the international levels to start green energy projects and industrial production systems. Most importantly, government should provide tax benefits for green energy industries to encourage potential investors. Fourth, the governments should examine their countries’ legal structure, property rights laws, domestic and international trade policies, and regulation of business and credit markets to ensure institutional quality. Finally, the empirical findings of this study endorse the need for cooperation among the EU countries to reduce CO2 emissions, exchange technological innovations, share sustainable development ideas, and ensure adequate financial resources.

Our findings are limited to EU countries. In addition, the study does not perform separate comparison analyses for low- and high-income EU countries. Future research could examine the overall global social response to reducing CO2 emissions and increasing clean energy consumption by comparing developing and developed economies across the world. Such an investigation would provide additional insights by identifying the determinants of reducing CO2 emissions and encouraging the use of clean energy sources. It may also support policymakers in their efforts to develop environmentally friendly policies that can lead to sustainable economic growth.