1 Introduction

It is well known that renewable energy represents an important source of energy because it is undamaging to the environment and public health. This type of energy is a continuous source that capable to increase the energy security of a country by reducing its dependency on fossil fuels. The European Union (EU) is the second biggest producers of renewable energy after the USA (World Development Indicators 2014). The role of non-carbon electricity generation in Europe increased substantially in the last two decades. In 2012, more than 55 % of total electricity in the Europe was generated from renewable energy (Energy Information Administration 2014). This phenomenon reflects the European countries’ efforts in reaching the greenhouse emission targets. Most of the renewable electricity production comes from the combustible renewables and waste, hydroelectric power, nuclear power, solar power, and wind power. Renewable electricity generation has a substantial share to the total production of electricity in the Europe, which can be seen in Fig. 1.

Fig. 1
figure 1

Electricity production based on source as a percentage of total electricity production in 2012

Full size image

The increase in renewable electricity production and the reduction in fossil fuels electricity production during the past 24 years might have a significant impact in reducing the pollution level in the 23Footnote 1 European Union countries, namely Austria, Belgium, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Luxembourg, the Netherlands, Norway, Poland, Portugal, Slovakia, Slovenia, Spain, Sweden, Switzerland, and the UK.

The relationship between pollution, economic activities, and energy consumption has been thoroughly investigated by different scholars. These studies are summarized in Table 1. Most of the previous studies utilized GDP growth (Soytas et al. 2007; Sadorsky 2009; Menyah and Wolde-Rufael 2010a, b; Lean and Smyth 2010; Acaravci and Ozturk 2010; Pao and Tsai 2010; Hossain 2011; Bloch et al. 2012; Jafari et al. 2012; Shahbaz et al. 2013a, b; Govindaraju and Tang 2013; Apergis and Payne 2014; Baek and Pride 2014; Al-Mulali et al. 2015; and so forth), energy consumption (Apergis and Payne 2009, 2010; Hossain 2011; Zhang and Cheng 2009; Bloch et al. 2012; Chandran and Tang 2013; Shahbaz et al. 2013a, b; Saboori and Sulaiman 2013a, b; Bella et al. 2014; Boutabba 2014; Begum et al. 2015; Alshehry and Belloumi 2015; and so forth), urbanization (Hossain 2011; Zhang and Cheng 2009; Jafari et al. 2012; Zhang and Lin 2012; Al-mulali 2014a, b; Zhang et al. 2014; Shafiei and Salim 2014; Kasman and Duman 2015), population (Zhang and Cheng 2009; Zhang and Lin 2012; Omri 2013; Apergis and Payne 2014; Shafiei and Salim 2014; Alam et al. 2014), trade openness (Halicioglu 2009; Hossain 2011; Jayanthakumaran et al. 2012; Al-mulali 2012; Shahbaz et al. 2013a, b; Kohler 2013; Farhani et al. 2014; Yang and Zhao 2014; Boutabba 2014; Sebri and Ben-Salha 2014; Al-mulali and Ozturk 2015; Shahbaz et al. 2015; Kasman and Duman 2015), and financial development (Al-mulali and Che Sab 2012a, b; Omri 2013; Shahbaz et al. 2013a, b; Ozturk and Acaravci 2013; Boutabba 2014; Alam et al. 2014; Ziaei 2015) as indicators of economic activities. Most of the studies revealed that the above economic indicators were the main sources of pollution as they had long-run and short-run significant impacts on pollution. Moreover, most of the studies utilized CO2 emission as an indicator of environmental pollution (Halicioglu 2009; Sadorsky 2009; Apergis and Payne 2010; Menyah and Wolde-Rufael 2010a, b; Zhang and Cheng 2009; Hossain 2011; Alam et al. 2012; Ozturk and Uddin 2012; Bloch et al. 2012; Jafari et al. 2012; Shahbaz et al. 2013a; Chandran and Tang 2013; Saboori and Sulaiman 2013a, b; Apergis and Payne 2014; Baek and Pride 2014; Bella et al. 2014; Boutabba 2014; Heidari et al. 2015; Kasman and Duman 2015; Zhang and Da 2015; and so forth).

Table 1 Summary of the literature review
Full size table

The influence of renewable energy on pollution was investigated (Sadorsky 2009; Menyah and Wolde-Rufael 2010a, b; Bengochea and Faet 2012; Apergis and Payne 2014; Baek and Pride 2014; Shafiei and Salim 2014; Bolük and Mert 2014; Farhani and Shahbaz 2014), but in rare cases especially in the European Union countries. In addition, most of the studies utilized total renewable energy consumption in general. However, disaggregated renewable energy consumption by source is important as it can provide an accurate relationship between each renewable energy consumption source and pollution. Moreover, each source of renewable energy might have different effects on pollution. Hence, this study examines the effect of renewable energy productionFootnote 2 on pollution in European countries because Europe is the second largest renewable energy producer in the world. Moreover, this study also examines the effect of five disaggregated renewable energy sources on pollution, respectively. This step will distinguish which renewable energy source is significant in reducing pollution so that more accurate policy recommendations could be made.

2 Data and methodology

Twenty three European countries namely Austria, Belgium, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Luxembourg, the Netherlands, Norway, Poland, Portugal, Slovakia, Slovenia, Spain, Sweden, Switzerland, and the UK are taken as the sample of this study. Learning from the literature, this study utilizes four variables that can influence pollution namely the gross domestic product (GDP) and trade of goods and services (TD) as an indicator of trade openness, and domestic credit to private sector (DC) as an indicator of financial development. These variables are measured in millions of constant 2005 US dollars. In addition, urban population measured in thousands of individuals was also utilized as an indicator of urbanization. This study utilizes five sources of renewable electricity production (ECP), namely combustible renewables and waste generation (ECRW), hydroelectric generation (ECH), nuclear generation (ECNU), solar generation (ECS), and wind-powered generation (ECW), which are all measured in millions of kilowatt hours. Lastly, carbon dioxide emission (CO2) was used as indicator of pollution measured in thousands of CO2 metric tons. Annual data for all variables mentioned above were retrieved from the Euromonitor database (2014) for the period of 1990–2013.

The main model is presented as follows:

$$ {\text{CO}}_{2} \, = f\left( {{\text{GDP}},{\text{ TD}},{\text{ UR}},{\text{ DC}},{\text{ ECP}}} \right) $$
(1)

Since each of the variables was presented in its natural logarithm and the error term is included in the panel models, the econometric models can be presented as follows:

$$ LCO_{2it} = \beta_{i0} + \beta_{1i} LGDP_{it} + \beta_{2i} LTD_{it} + \beta_{3i} LUR_{it} + \beta_{4i} DC_{it} + \beta_{5i} ECRW_{it} + \varepsilon_{it} $$
(2)
$$ LCO_{2it} = \beta_{i0} + \beta_{1i} LGDP_{it} + \beta_{2i} LTD_{it} + \beta_{3i} LUR_{it} + \beta_{4i} DC_{it} + \beta_{5i} ECH_{it} + \varepsilon_{it} $$
(3)
$$ LCO_{2it} = \beta_{i0} + \beta_{1i} LGDP_{it} + \beta_{2i} LTD_{it} + \beta_{3i} LUR_{it} + \beta_{4i} DC_{it} + \beta_{5i} ECNU_{it} + \varepsilon_{it} $$
(4)
$$ LCO_{2it} = \beta_{i0} + \beta_{1i} LGDP_{it} + \beta_{2i} LTD_{it} + \beta_{3i} LUR_{it} + \beta_{4i} DC_{it} + \beta_{5i} ECS_{it} + \varepsilon_{it} $$
(5)
$$ LCO2_{it} = \beta_{i0} + \beta_{1i} LGDP_{it} + \beta_{2i} LTD_{it} + \beta_{3i} LUR_{it} + \beta_{4i} DC_{it} + \beta_{5i} ECW_{it} + \varepsilon_{it} $$
(6)

The symbols β 1, β 2, β 3, β 4, and β 5 are the slop coefficients, t represents the time series (1990–2013), i is the cross sections (23 countries for model 1 and model 2, 13 countries for model 3, and 21 countries for model 4 and 5),Footnote 3 and ε represents the error term.

The econometric analysis begins with panel unit root test to examine the integration of each variable. For robustness, three types of panel unit root tests were utilized, namely the Im–Pesaran–Shin (IPS), proposed by Im et al. (2003), ADF-Fisher and PP-Fisher, proposed by Maddala and Wu (1999). The IPS unit root permits heterogeneity in the dynamics of the autoregressive coefficients, while the ADF-Fisher and the PP-Fisher unit root allows heterogeneity across panel units. The three above panel unit root tests work under the null hypothesis of a panel unit root (non-stationary variables) and the alternative hypothesis of no unit root (stationary variables).

After examining the panel unit root and the integration of the variables indicated to be in order one (stationary at the first difference), the next step was to examine whether a long-run relationship between the variables exists. Therefore, the panel cointegration test was implemented. This study used the Pedroni (1999, 2004) cointegration test which is based on the Engle and Granger (1987) cointegration test that explain whether the residual of each variable is stationary at level which means that the variables are cointegrated, or I(1) which indicates that the variables are not cointegrated. The Pedroni cointegration procedure contains several statistical tests between and within dimension to examine whether the null hypothesis of no cointegration can be rejected. The Pedroni cointegration test works under the following regression:

$$ y_{it} = \alpha_{i} + \delta_{i} t + \beta_{1i} x_{1i,t} + \beta_{2i} x_{2i,t} + \cdots + \beta_{Mi} x_{Mit} + e _{i,t} $$
(7)

y and x are presumed to be integrated in order (1), α i and δ i are the individual and trend effects, while e represents the residuals. If the residuals in regression (7) were integrated in order (1), the null hypothesis of no cointegration cannot be rejected. To examine the integration of the residual, one of the following regressions is used:

$$ e_{it} = \rho_{i} e_{it - 1} + u_{it} $$
(8)
$$ e_{it} = \rho_{i} e_{it - 1} + \mathop \sum \limits_{j = 1}^{{p_{i} }} \psi_{ij} \Delta e_{it - j} + v_{it} $$
(9)

Regressions 8 and 9 can be utilized for each cross section.

If cointegration is concluded among the variables, the panel-pooled fully modified ordinary least square (FMOLS) will be implemented to analyze the long-run cointegration relationship between the dependent and the independent variables. The pooled FMOLS was proposed by Phillips and Moon (1999). This cointegrating regression is more capable of preventing spurious regression generated from the involvement of the I(1) variables which can cause misleading results. The pooled FMOLS estimator is presented below:

$$ \hat{\beta }_{FP} = \left( {\mathop \sum \limits_{i = 1}^{N} \mathop \sum \limits_{t = 1}^{T} \tilde{X}_{it} \tilde{y}_{it} } \right)^{ - 1} \mathop \sum \limits_{i = 1}^{N} \mathop \sum \limits_{t = 1}^{T} \left( {\tilde{X}_{it} \tilde{y}_{it} - \hat{\lambda }^{ + }_{{12^{'} }} } \right) $$
(10)

\( \tilde{X}_{it} \tilde{y}_{it} \) are the corresponding data removed from the individual deterministic trends and λ represents the cointegration regressors. It is fundamental to note that the pooled FMOLS estimator sums across cross sections separately in the numerator and denominator.

If cointegration is confirmed among the variables, there might be a causal relationship between the variables, at least in one direction. Therefore, the Granger causality was utilized. If cointegration exists, then the Granger causality based on vector error correction model (VECM) will be used. The VECM Granger causality can capture the short-run causality based on the F-statistic and the long-run causality based on the lagged error correction term. The VECM Granger causality is presented below:

$$ \begin{aligned} \left[ {\begin{array}{*{20}c} {\Delta LCO_{2it} } \\ {\Delta LGDP_{it} } \\ {\Delta LTD_{it} } \\ {\Delta LUR_{it} } \\ {\Delta LDC_{it} } \\ {\Delta LECP_{it} } \\ \end{array} } \right] & = \left[ {\begin{array}{*{20}c} {\alpha_{1} } \\ {\alpha_{2} } \\ {\alpha_{3} } \\ {\alpha_{4} } \\ {\alpha_{5} } \\ {\alpha_{6} } \\ \end{array} } \right] + \sum\limits_{p = - 1}^{r} {\left[ {\begin{array}{*{20}c} {\beta_{11,p} } & {\beta_{12,p} } & {\beta_{13,p} } & {\beta_{14,p} } & {\beta_{15,p} } & {\beta_{16,p} } \\ {\beta_{21,p} } & {\beta_{22,p} } & {\beta_{23,p} } & {\beta_{24,p} } & {\beta_{25,p} } & {\beta_{26,p} } \\ {\beta_{31,p} } & {\beta_{32,p} } & {\beta_{33,p} } & {\beta_{34,p} } & {\beta_{35,p} } & {\beta_{36,p} } \\ {\beta_{41,p} } & {\beta_{42,p} } & {\beta_{43,p} } & {\beta_{44,p} } & {\beta_{45,p} } & {\beta_{46,p} } \\ {\beta_{51,p} } & {\beta_{52,p} } & {\beta_{53,p} } & {\beta_{54,p} } & {\beta_{55,p} } & {\beta_{56,p} } \\ {\beta_{61,p} } & {\beta_{62,p} } & {\beta_{63,p} } & {\beta_{64,p} } & {\beta_{65,p} } & {\beta_{66,p} } \\ \end{array} } \right]} \left[ {\begin{array}{*{20}c} {\Delta LCO2_{it - p} } \\ {\Delta LGDP_{it - p} } \\ {\Delta LTD_{it - p} } \\ {\Delta LUR_{it - p} } \\ {\Delta LDC_{it - p} } \\ {\Delta LECP_{it - p} } \\ \end{array} } \right] \\ & \quad + \left[ {\begin{array}{*{20}c} {\varphi_{1} } \\ {\varphi_{2} } \\ {\varphi_{3} } \\ {\varphi_{4} } \\ {\varphi_{5} } \\ {\varphi_{6} } \\ \end{array} } \right]ect_{it - 1} + \left[ {\begin{array}{*{20}c} {\varepsilon_{1it} } \\ {\varepsilon_{2it} } \\ {\varepsilon_{3it} } \\ {\varepsilon_{4it} } \\ {\varepsilon_{5it} } \\ {\varepsilon_{6it} } \\ \end{array} } \right] \\ \end{aligned} $$
(11)

However, if the variables are not cointegrated, the Granger causality based on vector autoregressive (VAR) model will be used. The VAR Granger causality can only show the long-run causality among the variables. The VAR Granger causality model is presented below:

$$ \begin{aligned} \left[ {\begin{array}{*{20}c} {\Delta LCO_{2it} } \\ {\Delta LGDP_{it} } \\ {\Delta LTD_{it} } \\ {\Delta LUR_{it} } \\ {\Delta LDC_{it} } \\ {\Delta LECP_{it} } \\ \end{array} } \right] & = \left[ {\begin{array}{*{20}c} {\alpha_{1} } \\ {\alpha_{2} } \\ {\alpha_{3} } \\ {\alpha_{4} } \\ {\alpha_{5} } \\ {\alpha_{6} } \\ \end{array} } \right] + \sum\limits_{p = - 1}^{r} {\left[ {\begin{array}{*{20}c} {\beta_{11,p} } & {\beta_{12,p} } & {\beta_{13,p} } & {\beta_{14,p} } & {\beta_{15,p} } & {\beta_{16,p} } \\ {\beta_{21,p} } & {\beta_{22,p} } & {\beta_{23,p} } & {\beta_{24,p} } & {\beta_{25,p} } & {\beta_{26,p} } \\ {\beta_{31,p} } & {\beta_{32,p} } & {\beta_{33,p} } & {\beta_{34,p} } & {\beta_{35,p} } & {\beta_{36,p} } \\ {\beta_{41,p} } & {\beta_{42,p} } & {\beta_{43,p} } & {\beta_{44,p} } & {\beta_{45,p} } & {\beta_{46,p} } \\ {\beta_{51,p} } & {\beta_{52,p} } & {\beta_{53,p} } & {\beta_{54,p} } & {\beta_{55,p} } & {\beta_{56,p} } \\ {\beta_{61,p} } & {\beta_{62,p} } & {\beta_{63,p} } & {\beta_{64,p} } & {\beta_{65,p} } & {\beta_{66,p} } \\ \end{array} } \right]} \left[ {\begin{array}{*{20}c} {\Delta LCO2_{it - p} } \\ {\Delta LGDP_{it - p} } \\ {\Delta LTD_{it - p} } \\ {\Delta LUR_{it - p} } \\ {\Delta LDC_{it - p} } \\ {\Delta LECP_{it - p} } \\ \end{array} } \right] \\ & \quad + \left[ {\begin{array}{*{20}c} {\varepsilon_{1it} } \\ {\varepsilon_{2it} } \\ {\varepsilon_{3it} } \\ {\varepsilon_{4it} } \\ {\varepsilon_{5it} } \\ {\varepsilon_{6it} } \\ \end{array} } \right] \\ \end{aligned} $$
(12)

The i represents the cross section (number of countries), t denotes the time, ε it is the error term, and ect is the lagged error correction term.

3 Empirical results

As mentioned earlier, the first step in the econometric analysis is to examine the stationarity of the variables. The three panel unit root tests, namely Im–Pesaran–Shin (IPS), ADF-Fisher Chi-square and the PP-Fisher Chi-square were conducted. The panel unit root tests results are displayed in Table 2. The results indicate that the null hypothesis of a panel unit root at level is not rejected by any variable. This shows that the variables are not stationary at level. However, the null hypothesis of the panel unit root is rejected at the first difference because all the variables are significant at the first difference.

Table 2 Panel unit root tests results
Full size table

Since the variables are stationary at the first difference, the second step is to examine the long-run relationship between the variables for the four models of this study (Eqs. 2–6). Therefore, the Pedroni cointegration test was conducted, and its results are reviewed in Table 3. The results reveal that four statistics are significant which, consequently, reject the null hypothesis of no cointegration for all the five models. Therefore, the long-run relationship between LCO 2 , LGDP, LTD, LUR, LDC, and LECP is confirmed. This results was consistent with the outcome of a number of previous studies that also found a long-relationship between CO2 emission and its main determinants (Menyah and Wolde-Rufael 2010a, b; Hossain 2011; Chandran and Tang 2013; Shahbaz et al. 2013a, b; Saboori and Sulaiman 2013a, b; Apergis and Payne 2009, 2010, 2014; Baek and Pride 2014; Bella et al. 2014; Al-mulali 2014a, b; Boutabba 2014; Sebri and Ben-Salha 2014 and so forth).

Table 3 The results of Pedroni’s cointegration tests
Full size table

After cointegration is confirmed among the variables in all models, the panel-pooled FMOLS was utilized to examine the positive as well as the negative long-run relationship between the independent and dependent variables. The panel FMOLS results are shown in Table 4.

Table 4 The results of panel FMOLS
Full size table

The FMOLS results reveal that, in all the models, GDP growth, urbanization, and financial development increase CO2 emission in the long run, while trade openness reduces CO2 emission. The increase in GDP growth by 1 % will increase CO2 emission by 0.41, 0.66, 0.027, 0.46, and 0.97 % respectively. Moreover, the increase in urbanization by 1 % will increase CO2 emission by 0.39, 0.54, 0.49, 0.27, and 0.26 % respectively. This outcome was consistent a number of previous studies (Halicioglu 2009; Apergis and Payne 2009; Ozturk and Acaravci 2010; Lean and Smyth 2010; Hossain 2011; Chandran and Tang 2013; Shahbaz et al. 2013a, b; Baek and Pride 2014; Bella et al. 2014; Al-mulali 2014a, b; Farhani, et al. 2014; Boutabba 2014; Kasman, and Duman 2015 and so forth). Furthermore, financial development increases pollution by its positive effect on CO2 emission in the long run. The increase in financial development by 1 % will increase CO2 emission by 0.05, 0.05, 0.09, 0.10, and 0.10 %, respectively, for each model. These results were similar to Boutabba (2014), while the results were not comparable to what was found by Shahbaz et al. (2013a, b). However, trade openness reduces pollution as it has a negative long-run effect on CO2 emission. A 1 % increase in trade openness will reduce CO2 emission in model 1, 2, 3, 4, and 5 by 0.20, 0.31, 0.20, 0.34, and 0.5 %, respectively. These results were similar to what was found by Shahbaz et al. (2013a, b), Farhani et al. (2014), and Kohler (2013). However, other scholars found a positive relationship between the two variables such as Halicioglu (2009), Al-mulali (2012), Al-mulali and Sheau-Ting (2014), and Kasman and Duman (2015).

The results for the long-run relationship between renewable electricity production by source and CO2 emission differs across the models. For model 1, the results show that electricity production from combustible renewables and waste generation has a significant negative long-run effect on CO2 emission. A 1 % increase in this source of electricity will reduce CO2 emission by 0.1 %. Similarly, electricity production from hydroelectric generation has a significant negative effect on CO2 emission as its increase by 1 % will reduce CO2 emission by 0.2 %. However, the increase in electricity production from nuclear generation, solar generation, and wind-powered generation has a negative relationship, but an insignificant effect on CO2 emission. A number of studies have also reached to the same results (Baek and Pride 2014; Shafiei and Salim 2014 Zeb et al. 2014), but other studies found that the relationship between the two variables was positive or insignificant (Al-mulali 2014a, b; Bolük and Mert 2014; Farhani and Shahbaz 2014; Al-mulali et al. 2015).

Since the variables are cointegrated for all models, the Granger causality based on the VECM was utilized. The results are presented in Table 5. The results for model 1 reveal the existence of long-run causality between CO2 emission, GDP growth, financial development, and electricity production from combustible renewables and waste generation. The short-run causality shows a bidirectional causal relationship between CO2 emission and GDP growth, CO2 emission and urbanization, CO2 emission and financial development, GDP growth and trade openness, GDP growth and financial development, electricity production from combustible renewables and waste generation and GDP growth, trade openness and electricity production from combustible renewables and waste generation, and trade openness and financial development. Moreover, a unidirectional causality was also found from GDP growth to urbanization, from electricity production from combustible renewables and waste generation to CO2 emission, trade openness to electricity production from combustible renewables and waste generation, financial development to urbanization, and from urbanization to electricity production from combustible renewables and waste generation.

Table 5 VECM Granger causality results
Full size table

For model 2, the Granger causality outcome reveals a long-run causal relationship between CO2 emission, GDP growth, trade openness, financial development, and electricity production from hydroelectric generation. The short-run causality reveals the presence of a bidirectional causality between CO2 emission and GDP growth, CO2 emission and trade openness, CO2 emission and financial development, CO2 emission and electricity production from hydroelectric generation, GDP growth and trade openness, GDP growth and urbanization, GDP growth and financial development, and trade openness and electricity production from hydroelectric generation. However, one-way causality was found from CO2 emission to urbanization, from GDP growth to electricity production from hydroelectric generation, and from financial development to trade openness.

The Granger causality for model 3 reveals the existence of long-run causality between CO2 emission, financial development, and GDP growth. The short-run causality results show a bidirectional causality between CO2 emission and financial development, GDP growth and trade openness, and between GDP growth and financial development. On the other hand, a one-way causality was concluded from GDP growth to electricity production from nuclear generation, from electricity production from nuclear generation to CO2 emission, CO2 emission to financial development, and from financial development to trade openness.

The Granger causality outcome for model 4 shows a long-run bidirectional causality between trade openness and urbanization. The short-run causality reveals a bidirectional causality between CO2 emission and GDP growth, GDP growth and trade openness, and GDP growth and financial development. One-way causality was also confirmed from CO2 emission to trade openness, GDP growth to urbanization, GDP growth to electricity production from wind generation, trade openness to electricity production from wind generation, financial development to urbanization, and from urbanization to electricity production from wind generation.

The granger causality results in model 5 shows bidirectional long-run causality between GDP growth and financial development. The results for the short-run causality reveals a bidirectional causality between CO2 emission and GDP growth, GDP growth and trade openness, GDP growth and financial development, and urbanization and electricity production from solar generation. Moreover, one-way causality was found from CO2 emission to financial development, from GDP growth to electricity production from solar generation, and from financial development to trade openness.

The causal relationship between renewable energy consumption and CO2 emission was also confirmed by scholars such as Menyah and Wolde-Rufael (2010a, b), Apergis and Payne (2014), Al-mulali (2014a, b), Shafiei and Salim (2014), and Farhani and Shahbaz (2014).

4 Discussion of results

The results from FMOLS show clearly that GDP growth, urbanization, and financial development are the main contributors to CO2 emission in the European countries. Increase in economic activities, which include consumption, investment, government purchases (the main components of GDP), increases the demand for energy, and thus an increase in electricity consumption. A share of electricity consumption comes from fossil fuels (fossil fuels represent 20 % of total electricity generation in Europe) which are the main sources of greenhouse gases. With better job opportunities, urban population in the European countries is substantially increasing to the point that in 2013 urban population represented over 50 % of total population. This percentage is expected to increase in the future. The increasing density of urban population will cause the deterioration of air quality due to, for instance, the increase in electricity consumption, automobiles, and the loss of tree cover as a result of urban development.

Furthermore, the domestic credit to the private sector increases CO2 emission in the long run. This phenomenon indicates that the financial resources that were provided for the private sectors are invested in non-environmental friendly projects. The trade openness reduces pollution in the long run in these countries, which indicates that the trade-related actions and strategies to increase environmental protection in these countries reached a point where it can reduce the environmental pollution induced by trade in general. Moreover, from the results, it seems that trade openness is stimulating the non-polluted industries which may explain the negative relationship between trade openness and CO2 emission. Furthermore, the results for renewable energy by source were diverse because, despite that all of these renewable sources have a negative effect on pollution, only three types of renewable energy sources were significant in reducing CO2 emission. Electricity production generated by combustible renewables and waste generation, hydroelectric generation, and nuclear generation was the only renewable energy source that reduces CO2 emission significantly. This outcome can be clarified by indicating that the share of these three sources of renewable energy plays a significant portion of total electricity production that 55 % of total electricity produced in 2013 comes from renewable electricity production (Euromonitor 2014). For this reason, electricity production that is generated by solar and wind energy has an insignificant effect on CO2 emission because the share of these sources in the total electricity production is small.

From the Granger causality results (focusing on the short-run causality), it was concluded that electricity production generated by combustible renewables and waste generation, hydroelectric generation, and nuclear generation was the only renewable energy source that has a negative causal relationship with CO2 emission. However, electricity production from wind and solar generation has no causal effects on CO2 emission. Moreover, GDP growth has a positive causal effect on CO2 emission in all models. This indicates that GDP growth is resulting in increasing CO2 emission in the short run. In spite of the outcome that urbanization and financial development have a positive causal effect on CO2 emission, this phenomenon was only confirmed in few cases. Furthermore, trade openness has a negative causality with CO2 emission in all the models except in model 2 where it was significant. Therefore, in most cases, urbanization, financial development, and trade openness have no significant causal effect in influencing CO2 emission in the short run. Regarding renewable electricity generation based on source, it is confirmed that GDP growth has a positive causal effect in influencing all renewable electricity sources in the short run.

5 Conclusion and policy implications

There is a lack of studies that investigated the effect of renewable energy by source on pollution in the Europe despite that over 55 % of its electricity is generated from renewable sources. Therefore, the researchers of this study were encouraged to examine the influence of electricity production from renewable generation on CO2 emission in 23 selected European countries. To achieve the study objectives, this research utilized the panel data techniques taking the period of 1990–2013. The outcome from the Pedroni cointegration indicated the existence of a long-run relationship between CO2 emission, GDP growth, urbanization, trade openness, financial development, and renewable electricity from all sources. In addition, the FMOLS results revealed that GDP growth, urbanization, and financial development are the main factors that influence CO2 emission positively while trade openness reduces CO2 emission in the long run. However, electricity production from combustible renewables and waste generation, hydroelectric generation, and nuclear generation was the only renewable source that influences CO2 emission significantly in the long run, while electricity production from wind and solar generation was insignificant.

Moreover, the VECM Granger causality showed that GDP growth is the most significant determinant that has positive causal effect on CO2 emission while urbanization and financial development have a positive causality on CO2 emission, but only in few of the investigated models. Moreover, trade openness has a significant negative causality on CO2 emission, but it was only verified in model 2. Furthermore, it was found that GDP is the main determinant that has a significant positive causal effect on all renewable electricity production, while trade openness and urbanization have a positive causal influences on renewable electricity production, but in only few models.

From the outcome of this study, a number of policy recommendations can be provided for the investigated countries. Since GDP, urbanization, and financial development increase CO2 emission, it is important to increase projects and investments that promote the role of renewable energy by providing incentives to the renewable manufactories and promoting new research in renewable energy technologies. This can increase the role of renewable energy which, as a result, will not only aid in creating more jobs in construction and manufacturing but will also help the renewable energy technologies to achieve economies of scale to reduce the cost of these sources of energy. Moreover, these countries should also increase their consumption of cleaner sources of fossil fuels, such as natural gas and higher-grade coal. In addition, encouraging the private sector to invest in more environmentally friendly projects and investments as well as increasing regulations that control the activities of the private sector can prevent the pollution caused by the credits or the financial resources that the banks provide to the private sector. Also, it is essential to utilize trade openness to stimulate non-polluted industries by inducing tax on polluted industries and establishing incentives on non-polluted industries to encourage producers to shift toward cleaner and more environmentally friendly industries. All these policies can help the countries to increase their energy efficiency in general which, consequently, reduces their environmental degradation that is caused by higher economic activities.