The Effects of Renewable Energy, Economic Growth, and Trade on CO₂ Emissions in the EU-15

Abstract

This study examines the impact of renewable energy, economic growth, and trade openness on CO₂ emissions in the EU-15 countries over the period 1980–2022, employing the ARDL modeling framework. In addition, a panel PMG-ARDL model is employed as a robustness check. The analysis identifies cointegration among the variables in 11 out of the 15 countries studied. Economic growth is found to increase CO₂ emissions, highlighting the ongoing challenge of aligning economic expansion with environmental objectives. The estimated coefficients for economic growth range from 0.43 to 5.70, depending on the country. Renewable energy significantly reduces emissions, highlighting its critical role in achieving sustainability (the corresponding coefficient moves in the range −0.13 to −0.96). Trade openness generally shows a neutral impact on emissions across most cases. Overall, renewable energy contributes to reducing CO₂ emissions, whereas the effects of economic growth and trade openness remain mixed and country-specific. These findings highlight the need to promote cleaner technologies, enhance energy efficiency, and ensure broader access to environmentally friendly energy sources.

1. Introduction

In 2015, United Nations member states adopted Agenda 2030, the most comprehensive action plan to date for promoting social inclusion, environmental sustainability, and economic growth. This agenda encompasses 17 Sustainable Development Goals, designed to guide development policies from 2015 to 2030. Growing concern about global warming and climate change among scientists, policymakers, and governments has brought significant attention to the issue of environmental pollution. According to the EIA [1], global CO₂ emissions related to energy reached nearly 37.8 billion metric tons in 2024 and are projected to exceed 40 billion metric tons by 2040. For comparison, emissions in 2010 were estimated at 31 billion metric tons.

EU countries have aligned their climate and energy objectives with the SDGs, aiming for a further reduction in greenhouse gas (GHG) emissions. These countries face adverse climate trends and increasing economic losses caused by climate change, with significant annual variability. The EU is committed to achieving an economy with net-zero GHG emissions by 2050, which entails a substantial reduction in emissions and a complete phase-out of coal-based technologies. All sectors are expected to reach net-zero emissions to meet this long-term objective. To achieve climate neutrality by 2050, the EU has also set intermediate targets. By 2030, it aims to reduce CO₂ emissions by at least 55% compared to 1990 levels. As of the end of 2023, emissions had decreased by 37%, while the EU economy grew by 60% over the same period.

The purpose of this paper is to examine the impact of renewable energy consumption, trade openness, and economic growth on CO₂ emissions, with a specific focus on EU-15 countries over the period 1980–2022. Aligned with the established research aim, this study poses two key research questions:

• Does the growth of economic activity predominantly lead to an increase in CO₂ emissions?
• To what extent does the increased utilization of renewable energy sources contribute to the reduction in CO₂ emissions?

Between 2009 and 2019, energy consumption increased by an average of 1.9% annually [2]. The energy sector is responsible for more than 75% of the EU’s emissions, making the expansion of renewable energy across all sectors essential to achieving emission reduction targets. The updated Renewable Energy Directive increases the EU’s mandatory renewable energy target for 2030 to at least 42.5%. In the second quarter of 2024, EU economic emissions were estimated at 790 million tonnes of CO₂ equivalents, representing a 2.6% decrease compared to the same quarter in 2023. Among EU-15 countries, Denmark, Spain, Portugal, Belgium, Italy, the Netherlands, France, and Germany were estimated to have reduced emissions while simultaneously growing their GDP.

Several prior studies have examined the effects of economic growth, renewable energy, and trade openness on CO₂ emissions. Reducing CO₂ emissions is often achieved through the adoption of clean energy sources and a circular economy [3,4]. On that path, Refs. [5,6,7] established that the role of renewable energy sources is of crucial importance. A similar result was observed in the 15 major renewable energy-consuming countries [8], and Visegrad countries [9,10] have determined that the status of economic growth does not automatically reduce climate vulnerability, while Refs. [11,12] showed that in nearly all countries, there was some degree of decoupling between GDP growth and CO₂ emissions, whether strong or weak. Ref. [13] established that the exploitation of natural resources and trade openness have a long-term negative impact on environmental quality, while from the perspective of correlation [14,15], it was shown that trade openness is positively correlated with CO₂ emissions.

Although research in this area has advanced significantly, notable gaps persist. These include regional disparities, limited exploration of long-term dynamics, the influence of timeframe selection, and the use of appropriate econometric methodologies. This study seeks to address some of these gaps through a more nuanced and methodologically rigorous approach. Analyses that combine time series and panel econometrics, particularly over extended timeframes and using updated data, remain relatively limited. This paper’s contribution is twofold. First, the subjects of the analysis are highly developed EU countries that should be the bearers of changes in the area of environmental protection, and their achievements will form future solutions to a great extent that will be applied by less developed countries. Secondly, methodologically, in this paper, by applying certain techniques, each country will be analyzed individually, having in mind the specificity of each national economy, contrary to most previous studies, which used just panel analysis. Additionally, as a special variable, GDP will be used as a closer indicator of economic growth, contrary to previous studies, which mainly used GDPpc.

This paper is structured as follows. Section 1 provides the background and motivation for this study. Section 2 examines the main determinants of CO₂ emissions. Section 3 and Section 4 outline the models employed and the data used. The empirical results are presented and discussed in Section 5. Finally, Section 6 summarizes the key findings and discusses their implications.

2. Literature Review

A significant body of empirical research has examined the comparative impacts of renewable and non-renewable energy, typically using panel data analysis or single-country case studies. In addition, scholars have explored the relationship between CO₂ emissions and various macroeconomic and structural determinants. This includes analyzing the links between carbon emissions and key variables such as gross domestic product (GDP), renewable energy consumption, and trade openness. The following sections present a comprehensive review of relevant studies, with a focus on the following: the interrelationship between GDP and CO₂ emissions, the role of renewable energy in mitigating emissions, and the influence of trade on carbon emissions. This review also emphasizes the importance of understanding these dynamics to support evidence-based policymaking and address the pressing challenges of climate change effectively.

2.1. The Relationship Between Economic Growth and CO₂ Emissions

While economic growth is commonly identified as a major driver of CO₂ emissions, its specific impact varies across countries [16]. This dynamic has attracted both academic and public attention, and it can be examined from two primary causal perspectives: from CO₂ emissions to economic growth [17] or from economic growth to CO₂ emissions [3,18]. The primary aim of this study is to analyze how economic expansion influences CO₂ emissions.

Previous research has identified six distinct patterns in the relationship between economic growth and CO₂ emissions [19]. The first corresponds to the well-known inverted U-shaped Environmental Kuznets Curve (EKC), which posits that CO₂ emissions initially increase with economic growth but begin to decline once a certain level of economic development is achieved. This trend reflects a transition from dependence on inexpensive hydrocarbon fuels during the early stages of growth to a greater reliance on renewable energy sources as living standards and environmental awareness improve. This pattern is supported by several studies, particularly in the context of EU countries [20,21,22].

The second observed relationship is a U-shaped curve [23], while the third follows an N-shaped pattern [24]. Evidence of an N-shaped relationship between emissions and economic growth has been found in a sample of the so-called EU-5 countries [25]. The fourth pattern resembles an inverted N-shaped curve [26,27]; similarly, an inverted N-shaped relationship between economic growth and emissions has been observed in the EU-28 [28]. The fifth category suggests that GDP growth contributes to a reduction in CO₂ emissions, as demonstrated in studies such as [13,16,29]. In contrast, the sixth category indicates that economic growth leads to increased CO₂ emissions [30,31].

In many developing countries, economic expansion is closely linked to increased energy demand, often met through fossil fuels, which intensifies CO₂ emissions [32,33]. A comprehensive meta-analysis, covering the period from 1995 to 2017, highlights that efforts to curb CO₂ emissions often involve trade-offs with economic performance, emphasizing the complex challenge of aligning environmental sustainability with economic development [20].

2.2. The Relationship Between Renewable Energy and CO₂ Emissions

The relationship between energy production and CO₂ emissions indicates that non-renewable energy sources tend to increase emissions, while renewable energy sources contribute to their reduction [34,35,36]. However, the effects of renewable energy consumption on economic development and environmental outcomes are often context-dependent. For instance, Ref. [35] emphasized that the impact of renewable energy on economic growth varies according to the level of risk within a given economy. In a related area of research, some authors investigated the relationship between tourism and environmental sustainability, using the European Union as a case study [15,37]. Their empirical analysis revealed a negative correlation between tourist arrivals and CO₂ emissions, indicating that the tourism sector can contribute to environmental objectives when sustainable practices are implemented. This finding highlights the potential for tourism to simultaneously foster economic growth and environmental stewardship.

Shahnazi and Shabani demonstrated that the use of renewable energy in the EU contributes to a reduction in CO₂ emissions [19]. However, empirical evidence on this relationship remains mixed across countries and time periods. For instance, Rahman and Vu found that renewable energy significantly reduces emissions in both the short and long run in Australia but reported no significant effect in Canada [22]. Research on the BRICS economies suggests that renewable energy had an insignificant impact on emissions between 1992 and 2016, providing limited insight into the observed changes in CO₂ levels [38]. Contrasting results have also emerged in other regions. It was found that renewable energy paradoxically increased CO₂ emissions, possibly due to transitional inefficiencies or reliance on fossil fuel-based backup systems [39]. Recommendations for Egypt emphasize pursuing equitable growth through sustainable energy innovation, attracting foreign direct investment (FDI), and fostering long-term economic development [40].

Region-specific studies further highlight the complexity of the relationship between renewable energy and CO₂ emissions. For example, renewable energy was found to reduce emissions in China’s western and eastern regions, while the effect was statistically insignificant in the central region [41]. Similarly, a strong negative relationship between renewable energy and CO₂ emissions has been identified across a group of major economies—including China, the United States, Japan, Canada, Brazil, South Korea, and Germany—particularly at the tails of the distribution [42]. Evidence from Turkey over 1980–2016 suggests that renewable energy consumption does not have a statistically significant effect on emissions [43]. The impact of renewable energy in Italy has been found to vary seasonally, offering further nuance to the analysis [44]. During summer months, emission reductions were more pronounced, driven by the increased use of Combined Cycle Gas Turbines (CCGTs) and a corresponding decline in coal-based power generation. Moreover, their analysis revealed that in the Italian energy system, 1 kWh of renewable energy displaced only about 0.8 kWh of conventional energy—highlighting the partial substitution effect of renewables and the continued dependence on traditional sources.

2.3. The Relationship Between Trade and CO₂ Emissions

Theoretically, the impact of trade on CO₂ emissions can be either positive or negative, depending on a country’s economic structure, regulatory framework, and access to technology. A positive environmental effect of trade may arise when greater integration into international markets enhances competitiveness and facilitates the import of cleaner, more efficient technologies, thereby contributing to a reduction in CO₂ emissions [45]. Conversely, trade can also have a negative environmental impact, especially when it fuels increased industrial activity and production, leading to a rise in CO₂ emissions and the degradation of environmental quality.

Several empirical studies support the view that trade expansion contributes to higher emissions. For example, some authors find a positive association between trade growth and CO₂ emissions [46,47,48,49,50]. Ref. [46], using a sample of 105 countries across different income groups, showed that trade openness has a detrimental effect on environmental quality, contributing to environmental degradation regardless of income level [51]. Other studies also report that greater trade intensity is correlated with higher emissions [52,53].

However, the findings from other researchers suggest more context-dependent or mixed outcomes. For instance, Ref. [22] found divergent results in Canada and Australia. In Canada, trade was associated with higher CO₂ emissions (short-run coefficient: 0.47), whereas in Australia, trade had a mitigating effect on emissions (short-run coefficient: –0.47). In a similar vein, a study analyzing Turkey identified a weak but negative relationship between trade and CO₂ emissions, suggesting that trade may offer modest environmental benefits [54]. Using Italy as a case study, Ref. [55] found that trade increased CO₂ emissions in the long run, although its short-term effect was not statistically significant. Further nuance is provided by [56], which disaggregated trade components and showed that exports tend to reduce emissions, while imports contribute to an increase—underscoring the importance of trade composition in shaping environmental outcomes

Contrasting these results, other authors reported no statistically significant relationship between trade and CO₂ emissions [57,58]. These findings emphasize the heterogeneous nature of the trade–emissions nexus, suggesting that its impact may vary substantially based on country-specific factors, institutional quality, energy mix, or the levels of technological advancement.

2.4. Research Gaps in the Literature

Typical conclusions in the emissions–growth–renewable energy literature emphasize several key insights. Chief among them is the recognition that increasing the share of renewable energy within the energy mix is essential for reducing CO₂ emissions without compromising economic growth. Additionally, the relationship between economic activity and emissions frequently aligns with the Environmental Kuznets Curve (EKC) hypothesis, which posits that emissions rise during early stages of growth but decline as economies adopt cleaner technologies and environmental awareness increases.

Specifically, this research aims to contribute to the formulation of sustainable energy transition policies in the EU by examining both time series and panel data. Through the application of econometric techniques—such as unit root tests, cointegration analysis, and panel cointegration methods—this study provides robust empirical evidence to inform policymaking. Given the EU-15’s ongoing commitment to emission reduction, energy efficiency, and the expansion of renewables in alignment with EU sustainability goals, the findings of this study are highly relevant.

By identifying key trends and anticipating future challenges—such as the phase-out of coal and the adoption of emerging technologies—this research aims to support the development of targeted and effective energy policies. The motivation behind this study lies in the ambition to deepen the understanding of the complex interplay between economic growth, environmental sustainability, and emissions, ultimately offering valuable evidence for informed, forward-looking decision-making.

3. Methods

Figure 1 presents the research framework, outlining the key concepts, relationships, and scope of this study.

3.1. Model and Theoretical Basis

In this paper, we examine the effects of renewable energy sources, economic growth, and trade openness on CO₂ emissions, using the example of the EU-15 countries over the period 1980–2022. The empirical relationship between economic growth and CO₂ emissions is not fully established, making it a suitable topic for further research. Energy can be viewed as an important factor of production. It is assumed that fossil fuel-based energy harms environmental quality, while energy derived from renewable sources significantly reduces CO₂ emissions. The impact of renewable energy on CO₂ emissions will be analyzed in this study. Trade openness is included as an additional variable in the model, as its influence on emissions can be important when shaping a country’s trade policy. In this context, the functional dependence between the variables can be represented by the following functional form:

$${{CO}_2}_t=f({GDP}_t,{RE}_t,{TO}_t)$$

(1)

The variables will be transformed into their logarithmic form in order to obtain coefficients of direct elasticity. The relationship can be represented in the following functional form:

$${ln{CO}_2}_t={\alpha }_0+{\alpha }_1{lnGDP}_t+{\alpha }_2{lnRE}_t+{\alpha }_3{lnTO}_t+{\mu }_t$$

(2)

where CO_2t represents carbon emission, GDP_t is the gross domestic product, RE_t denotes renewable energy, TO_t stands for trade openness, and μ_t is the error term.

3.2. Unit Root Test

An essential assumption in time series analysis involves testing whether the variables contain a unit root. In this paper, the ADF test is applied. The ADF test is based on the following regression equation:

$${\Delta X}_t={\psi }_0+{\psi }_1\times t+{\psi }_2{X}_{t-1}+{\sum }_{i=1}^k{\psi }_i{\Delta X}_{t-i}+{\epsilon }_t$$

(3)

where ψ₀ represents a constant, t is the time trend, and k is the selected lag length. The null hypothesis H₀: ψ₂ = 0 can be tested against the alternative hypothesis: ψ₂ < 0. When choosing deterministic components, the Stock and Watson test was used. More details can be found in [59].

3.3. ARDL Cointegration

This study employs the ARDL bound testing approach, developed by [60]. One of the key advantages of this method is its flexibility—it allows for the inclusion of variables that are integrated at different levels, i.e., I(0) or I(1), provided that none are integrated of order I(2) or higher. This makes the ARDL approach particularly suitable for small samples and models involving endogenous regressors. The model includes the lags of both dependent and independent variables, allowing for dynamic interactions. Each variable can have its own optimal lag length, selected based on appropriate criteria. However, if any variable is found to be integrated of order I(2) or higher, the ARDL model becomes invalid. To test for the presence of a long-run relationship, the bound testing procedure compares the calculated F-statistic to critical values. These critical values are provided in reference tables in [60,61,62].

The ARDL model assumes a linear functional form to describe the long-term relationship among the included variables. This approach offers a robust framework for analyzing dynamic relationships, even in the presence of heterogeneous dynamics and limited sample sizes:

$$\begin{array}{c}{\Delta \mathrm{l}\mathrm{n}{\mathrm{C}\mathrm{O}}_2}_{\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{d}\mathrm{u}\mathrm{m}}_{\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{l}\mathrm{n}{\mathrm{C}\mathrm{O}}_2}_{\mathrm{t}-1}+{\mathsf{\alpha }}_3{\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{t}-1}+{\mathsf{\alpha }}_4{\mathrm{l}\mathrm{n}\mathrm{R}\mathrm{E}}_{\mathrm{t}-1}+\hfill \\ {\mathsf{\alpha }}_5{\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{O}}_{\mathrm{t}-1}+{\sum }_{\mathrm{j}=0}^{\mathrm{q}}{\mathsf{\alpha }}_{\mathrm{j}}{\Delta \mathrm{l}\mathrm{n}{\mathrm{C}\mathrm{O}}_2}_{\mathrm{t}-\mathrm{j}}+{\sum }_{\mathrm{l}=0}^{\mathrm{m}}{\mathsf{\alpha }}_{\mathrm{l}}{\Delta \mathrm{l}\mathrm{n}\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{t}-\mathrm{l}}+{\sum }_{\mathrm{k}=0}^{\mathrm{n}}{\mathsf{\alpha }}_{\mathrm{k}}{\Delta \mathrm{l}\mathrm{n}\mathrm{R}\mathrm{E}}_{\mathrm{t}-\mathrm{k}}+\hfill \\ {\sum }_{\mathrm{n}=0}^{\mathrm{s}}{\mathsf{\alpha }}_{\mathrm{n}}{\Delta \mathrm{l}\mathrm{n}\mathrm{T}\mathrm{O}}_{\mathrm{t}-\mathrm{n}}+{\epsilon }_{\mathrm{t}}\hfill \end{array}$$

(4)

where ∆ denotes the first difference operator, dum represents a dummy variable for the structural break point, and ɛ is the error term assumed to be white noise. The coefficients α₂, α_3, α_4, and α₅ represent the long-run relationships, while α_j, α_l, α_k, and α_n are short-run coefficients. Dummy variables are exogenously determined and included to account for potential structural breaks, thereby improving the reliability and fit of the estimated model. The optimal lag length for the ARDL model is selected using the Akaike Information Criterion (AIC). Among all considered lag structures, the one with the lowest AIC value is chosen as optimal. The bound testing procedure is based on the joint F statistic that tests the null hypothesis that there is no cointegration in relation to the alternative that cointegration is present. The null hypothesis of no cointegration between the variables can be expressed as: H_o: α₂ = α₃ = α₄ = α₅ = 0. Not accepting the null hypothesis represents the confirmation of existing cointegration in the form of the alternative hypothesis H_a: α₂ ≠ α₃ ≠ α₄ ≠ α₅ ≠ 0.

Two sets of critical values are used for comparison in the bound testing procedure. The calculated F-statistic is evaluated against these bounds to determine the presence of cointegration. If the calculated F-statistic is greater than the upper bound, the null hypothesis of no cointegration is rejected, indicating a long-run relationship among the variables. Conversely, if the F-statistic is below the lower bound, the null hypothesis cannot be rejected, implying no cointegration. When the F-statistic falls between the lower and upper bounds, the outcome is inconclusive—unless the integration order of the variables is definitively known.

Since this study is based on a relatively small sample (43 observations), it is inappropriate to use the critical values from [60], which are designed for large samples. Therefore, to ensure the greater precision and reliability of inference, the critical values developed by [61] for small sample sizes are used. Once cointegration is confirmed, the second step involves estimating both the short-run and long-run dynamics of the model, based on the established long-term relationship among the variables.

$$\begin{array}{l}{\Delta ln{\mathrm{C}\mathrm{O}}_2}_t={\alpha }_0+{\alpha }_1{\Delta dum}_t+\sum _{\mathrm{j}=0}^{\mathrm{q}}{\mathsf{\alpha }}_{\mathrm{j}}{\Delta \mathrm{l}\mathrm{n}{\mathrm{C}\mathrm{O}}_2}_{\mathrm{t}-\mathrm{j}}+\sum _{\mathrm{l}=0}^{\mathrm{m}}{\mathsf{\alpha }}_{\mathrm{l}}{\Delta \mathrm{l}\mathrm{n}\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{t}-\mathrm{l}}+\\ \sum _{\mathrm{k}=0}^{\mathrm{n}}{\mathsf{\alpha }}_{\mathrm{k}}{\Delta \mathrm{l}\mathrm{n}\mathrm{R}\mathrm{E}}_{\mathrm{t}-\mathrm{k}}+\sum _{\mathrm{n}=0}^{\mathrm{s}}{\mathsf{\alpha }}_{\mathrm{n}}{\Delta \mathrm{l}\mathrm{n}\mathrm{T}\mathrm{O}}_{\mathrm{t}-\mathrm{n}}+{\beta ecm}_{t-1}+{\epsilon }_{\mathrm{t}}\end{array}$$

(5)

The coefficient of the error correction term (ECM), denoted as β, reflects the speed at which the variables return to equilibrium after a short-term disturbance. This coefficient must be negative and statistically significant to confirm the existence of a stable long-run relationship. To assess the adequacy of the estimated ARDL model, several diagnostic tests are conducted. These include tests for serial correlation, normality of the error term, and heteroscedasticity. The goal of these tests is to ensure the validity and reliability of the model’s estimates. For more detailed discussions on the application and interpretation of the ARDL approach and its diagnostics, see [63,64].

3.4. Panel Cointegration

To examine the robustness of the obtained results, a balanced panel dataset was constructed from the original data. To conserve space, the results of the panel unit root tests and the cross-sectional dependence tests are not presented in detail; instead, only the key findings are reported. Panel cointegration among the variables was tested using the method in [65]. Following this, the Pooled Mean Group (PMG) estimator within the ARDL framework was applied. The PMG-ARDL approach is particularly suited for panel data analysis, as it extends the traditional ARDL model to account for both short-term heterogeneity and long-term homogeneity across cross-sectional units. Specifically, the PMG method allows the following elements to differ across units:

• Intercepts: Accounting for individual-specific effects;
• Short-run dynamics: Capturing heterogeneity in the adjustment process;
• Cointegrating vector: Enforcing homogeneity in the long-run relationship across the panel.

The PMG model is particularly appropriate in contexts where slope heterogeneity is present, as noted by [66]. Its flexibility makes it well-suited for analyzing economic systems that exhibit common long-term trends while also allowing for distinct short-term fluctuations across countries, regions, or sectors. The PMG model can be expressed in the following general form:

$${\Delta Y}_{it}={\sum }_{j=0}^{m-1}{\Omega }_{i,t}{\Delta X}_{i,t-j}+{\sum }_{j=1}^{s-1}{\alpha }_{i,j}{\Delta Y}_{i,t-j}+{\beta ecm}_{i,t}+{\epsilon }_{i,t}$$

(6)

In this model, Δ denotes the difference operator. The dependent variable, Y, is influenced by the independent variables, X. To capture deviations from equilibrium, the adjustment coefficient (β) is incorporated. The error correction term (ecm_i,t) accounts for adjustments specific to each country (i) at time (t), while ε_it represents random errors or unexplained variations in the model.

4. Data

This study analyzes annual data from 1980 to 2022 for fifteen countries: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden, and the United Kingdom. Data on renewable energy (RE)—measured as electricity generation from renewable sources (in kWh)—and CO₂ emissions (in million metric tonnes) were obtained from the U.S. Energy Information Administration [1]. The chosen proxy for renewable energy was also employed by [67], while the CO₂ emissions measure aligns with that used in [68].

While most existing studies focus on the impact of renewable energy consumption on dependent variables such as CO₂ emissions or economic growth, this study specifically investigates the effect of renewable electricity generation on CO₂ emissions. Data for GDP (measured in constant 2015 US dollars) and trade openness (TO, defined as the sum of exports and imports of goods and services, also in constant 2015 US dollars) were sourced from the World Development Indicators [69]. These macroeconomic indicators have been utilized in previous empirical studies [70,71].

All variables were transformed into natural logarithmic form (ln). The trends in the logarithmic variables by country are illustrated in Appendix A. It should be noted that the variables lnGDP, lnRE, and lnTO are plotted on the left vertical axis, while lnCO₂ is displayed on the right vertical axis.

5. Results and Analysis

5.1. The Findings of Unit Root Tests

The results of the Augmented Dickey–Fuller (ADF) unit root test are presented in

Table 1. ADF unit root test results.

Variables	Austria		Belgium		Denmark		Finland		France
	At Level	At 1st Difference	At Level	At 1st Difference	At Level	At 1st Difference	At Level	At 1st Difference	At Level	At 1st Difference
	T-Stat	T-Stat	T-Stat	T-Stat	T-Stat	T-Stat	T-Stat	T-Stat	T-Stat	T-Stat
lnCO2t	−1.43	−6.43 **	−1.55	−8.19 **	0.61	−8.02 **	−0.12	−5.78 **	−0.84	−7.57 **
lnGDPt	−4.06 **	−10.71 **	−5.78 **	−5.58 **	−5.58 **	−9.63 **	−4.90 **	−9.99 **	−6.96 **	−6.26 **
lnREt	−1.69	−8.89 **	−1.10	−5.07 **	−3.89 **	−2.95 *	−1.37	−8.92 **	−1.34	−8.37 **
lnTOt	−2.89	−9.99 **	3.81 **	−6.53 **	−5.02 **	−6.27 **	−4.27 **	−10.16 **	−6.31	−5.93 **
Variables	Germany		Greece		Ireland		Italy		Luxembourg
	At level	At 1st difference	At level	At 1st difference	At level	At 1st difference	At level	At 1st difference	At level	At 1st difference
	t-stat	t-stat	t-stat	t-stat	t-stat	t-stat	t-stat	t-stat	t-stat	t-stat
lnCO2t	−0.26	−5.86 **	−1.67	−4.76 **	−1.68	−4.85 **	−0.79	−6.25 **	−2.05	−5.50 **
lnGDPt	−5.13 **	−10.67 **	−1.22	−2.27	−1.33	−7.28 **	−6.56 **	−6.10 **	−2.02	−5.09 **
lnREt	−1.41	4.85 **	0.46	−6.12 **	−0.86	−8.27	−0.37	−6.32 **	−1.78	−7.18 **
lnTOt	−3.12 *	−9.81 **	−5.51 **	−10.48 **	−1.18	−6.68 **	−5.71 **	−5.83 **	−0.62	−5.18 **
Variables	The Netherlands		Portugal		Spain		Sweden		United Kingdom
	At level	At 1st difference	At level	At 1st difference	At level	At 1st difference	At level	At 1st difference	At level	At 1st difference
	t-stat	t-stat	t-stat	t-stat	t-stat	t-stat	t-stat	t-stat	t-stat	t-stat
lnCO2t	−1.52	−2.89 *	−2.75	−7.23 **	−2.06	−5.41 **	−1.13	−6.85 **	0.97	−6.42 **
lnGDPt	−6.81 **	−5.85 **	−1.80	−2.93	−2.29	−2.80	−0.10	−5.72 **	−5.44 **	−7.00 *
lnREt	−0.88	−8.25 **	−0.77	−9.92 **	0.26	−8.13 **	−1.72	−9.42 **	1.30	−6.01 **
lnTOt	−5.83 **	−5.82 **	−2.84	−5.59 **	−1.71	−5.43 **	−0.48	−4.98 **	−4.76 **	−8.29 **

Notes: * and ** denote significance at 5% and 1%, respectively. Source: Authors’ calculations.

. The findings indicate that the variable lnGDP for Greece, Portugal, and Spain remains non-stationary even after first differencing, suggesting that it is integrated of order two, I(2). Given that the ARDL bound testing approach requires all variables to be either I(0) or I(1), further analysis for Greece, Portugal, and Spain was excluded from this study. For the remaining countries, the variables become stationary after first differencing (some are even stationary at level), with statistical significance at the 1% or 5% level.

This satisfies the preconditions for applying the ARDL modeling framework. The other analyzed variables across the remaining countries display a consistent pattern of non-stationarity at level, but they attain stationarity after transformation to first differences, confirming their suitability for further time series analysis.

5.2. The Findings of ARDL Tests

ARDL modeling is applied to examine the long-term relationship between lnCO₂ (as the dependent variable) and lnGDP, lnRE, and lnTO (as independent variables) for a panel of 12 EU countries.

Table 2. The findings of ARDL cointegration tests.

Estimated Model Flnco2 (lnCO2/lnGDP, lnRE, lnTO)
Country	Optimal Lag Length	F-Statistics
Austria	2, 1, 3, 3	6.68 **
Belgium	3, 1, 0, 1	8.46 ***
Denmark	2, 2, 2, 0	4.97 *
Finland	3, 1, 3, 1	7.73 ***
France	1, 1, 3, 1	4.87 *
Germany	3, 0, 0, 0	10.41 ***
Ireland	1, 0, 2, 0	9.17 ***
Italy	1, 3, 2, 3	21.81 ***
Luxembourg	2, 3, 0, 3	8.56 ***
The Netherlands	2, 0, 2, 0	8.63 ***
Sweden	1, 2, 0, 0	1.96
UK	4, 4, 0, 4	5.9 **
Case III
Significance level	I(0)	I(1)
10%	2.933	4.020
5%	3.548	4.803
1%	5.018	6.610
Case V
Significance level	I(0)	I(1)
10%	3.760	4.795
5%	4.510	5.643
1%	6.238	7.740

Notes: *, **, and *** denote significance at 10%, 5%, and 1%, respectively. Source: Authors’ calculations.

shows that in the cases of Belgium, Finland, Germany, Ireland, Italy, Luxembourg, and the Netherlands, the calculated F-statistic exceeds the critical values at the 1% significance level, indicating strong evidence of cointegration. For Austria and the United Kingdom, the F-statistic exceeds the critical values at the 5% level, while in the cases of Denmark and France, the threshold is surpassed at the 10% level, suggesting weaker—but still statistically significant—evidence of cointegration.

In contrast, for Sweden, no cointegration relationship is found among the observed variables, as the F-statistic does not exceed the lower bound of the critical values. Based on the dynamics of the variables (see Appendix A), Case III (unrestricted intercept, no trend) is used for the following countries: Austria, Belgium, Denmark, Finland, France, Ireland, Italy, Luxembourg, The Netherlands, and Sweden. Case V (unrestricted intercept and unrestricted trend) is applied for Germany and the United Kingdom.

5.3. The Long-Run and Short-Run Analyses

By testing the long-term relationships among the variables, this study examines the marginal effects of economic growth, renewable energy sources, and trade openness on CO₂ emissions. The results, presented in

Table 3. Long-run analysis.

Dependent Variable lnCO2t	Austria		Belgium		Denmark		Finland
Variables	Coeff	t-stat	Coeff	t-stat	Coeff	t-stat	Coeff	t-stat
lnGDPt	1.87	3.17 **	1.65	10.82 **	5.70	2.97 **	1.14	2.01 *
lnREt	−0.96	−3.44 **	−0.13	−6.51 **	−0.29	−2.61 **	−0.68	−1.15
lnTOt	−1.51	−0.53	−0.53	−3.99 **	−1.55	−3.49 **	−0.49	−1.65
Dependent variable lnCO2t	France		Germany		Ireland		Italy
Variables	Coeff	t-stat	Coeff	t-stat	Coeff	t-stat	Coeff	t-stat
lnGDPt	−2.20	−0.81	0.01	0.08	0.43	2.81 **	0.58	4.66 **
lnREt	−1.64	−1.19	0.27	3.24 **	−0.44	−2.14 *	−0.59	−4.03 **
lnTOt	2.05	0.92	−0.03	−0.34	0.20	1.91 *	0.66	2.29 **
Dependent variable lnCO2t	Luxembourg		The Netherlands		United Kingdom
Variables	Coeff	t-stat	Coeff	t-stat	Coeff	t-stat
lnGDPt	−2.61	−3.03 **	1.60	15.20 **	0.93	2.59 **
lnREt	−0.55	−2.58 **	0.01	0.06	−0.16	−2.63 **
lnTOt	2.14	3.13 **	−0.82	−15.61 **	0.67	1.92 *

Notes: * and ** denote significance at 5% and 1%, respectively. Source: Authors’ calculations.

, indicate that the coefficient of economic growth is positive and statistically significant in Austria, Belgium, Denmark, Finland, Ireland, Italy, the Netherlands, and the United Kingdom, suggesting that economic growth contributes to increased CO₂ emissions—consistent with prior findings [6,72,73]. The magnitude of the growth coefficient varies across countries, ranging from 0.43 in Ireland to 5.70 in Denmark. In Germany, the effect is also positive but not statistically significant. In France, economic growth appears to reduce CO₂ emissions, though the result lacks statistical significance. For Luxembourg, the coefficient indicates that a 1% increase in GDP reduces CO₂ emissions by 2.61%, aligning with the findings of [7,18].

Regarding renewable energy, Table 3 shows that in 7 out of 11 countries, the increased use of renewable sources significantly reduces CO₂ emissions, consistent with prior findings [19,74,75]. This relationship is observed in Austria, Belgium, Denmark, Ireland, Italy, Luxembourg, and the United Kingdom, with coefficients ranging from −0.13 in Belgium to −0.96 in Austria. Although France and Finland also show negative coefficients, these are not statistically significant, in line with [76]. In the Netherlands, the coefficient is positive but insignificant. Interestingly, in Germany, the model suggests that a 1% increase in renewable energy usage is associated with a 0.27% increase in CO₂ emissions, a result consistent with [39], possibly due to transitional inefficiencies or reliance on hybrid energy systems. In terms of trade openness, contrary to the findings by [45], the long-run effect is not statistically significant in 5 out of 11 countries, in line with [57,58]. Contrary to the findings of [46,47,48], trade openness is found to reduce CO₂ emissions in Belgium (−0.53), Denmark (−1.55), and the Netherlands (−0.82). In contrast, it is associated with an increase in emissions in Ireland (0.20), Luxembourg (2.14), and the United Kingdom (0.67), in line with [77].

In the short run, the effect of economic growth on CO₂ emissions is not statistically significant in Austria, Belgium, Denmark, Finland, France, or Luxembourg. As shown in

Table 4. Short-run analyses.

Dependent Variable ΔlnCO2t	Austria		Belgium		Denmark		Finland
Variables	Coeff	t-stat	Coeff	t-stat	Coeff	t-stat	Coeff	t-stat
ΔlnGDPt	−1.14	−1.56	0.27	0.57	1.50	1.70	0.48	0.96
ΔlnREt	0.55	3.82 **	−0.04	−0.85	−0.32	3.00 **	−0.19	−2.39 **
ΔlnTOt	0.53	2.99 **	0.08	0.78	−0.20	−0.83	0.21	1.18
ECMt−1	−0.41	−5.46 **	−1.04	−6.92 **	−0.60	−4.58 **	−0.46	−6.01 **
Adjusted R2	0.909		0.908		0.905		0.914
Dependent variable ΔlnCO2t	France		Germany		Ireland		Italy
Variables	Coeff	t-stat	Coeff	t-stat	Coeff	t-stat	Coeff	t-stat
ΔlnGDPt	−0.54	−1.43	0.72	3.26 **	−0.13	−2.47 **	0.57	2.28 **
ΔlnREt	−0.08	−3.10 **	−0.14	−2.99 **	−0.02	−0.52	−0.16	−5.64 **
ΔlnTOt	0.17	2.29 **	0.02	0.33	−0.02	−0.19	0.05	0.94
Dummy1991	-	-	−0.10	−5.59 **	-	-	-	-
ECMt−1	−0.11	−4.43 **	−0.46	−6.77 **	−0.70	−6.76 **	−0.64	−9.81 **
Adjusted R2	0.914		0.919		0.894		0.893
Dependent variable ΔlnCO2t	Luxembourg		The Netherlands		United Kingdom
Variables	Coeff	t-stat	Coeff	t-stat	Coeff	t-stat
ΔlnGDPt	−0.50	−1.14	1.46	4.08 *	0.77	3.22 **
ΔlnREt	0.37	3.64 *	0.001	0.06	0.10	2.47 **
ΔlnTOt	0.72	3.59 *	−0.33	−2.98 *	0.07	0.83
Dummy1997	-	-	-	-	−0.04	−2.14 *
Dummy 2002	-	-	-	-	−0.06	−3.13 **
ECMt−1	−0.33	−6.21 *	−1.12	−12.61 *	−0.36	−5.20 **
Adjusted R2	0.871		0.879		0.897

Notes: * and ** denote significance at the 5% and 1% level. Source: Authors’ calculations.

, in Germany, Italy, the Netherlands, and the United Kingdom, economic growth increases CO₂ emissions, with corresponding coefficients of 0.72, 0.57, 1.46, and 0.77, respectively. In Ireland, however, a 1% increase in economic growth is associated with a 0.13% reduction in CO₂ emissions.

The growth of renewable energy sources significantly reduces CO₂ emissions in Denmark, Finland, France, Germany, and Italy, with coefficients ranging from −0.08 (France) to −0.32 (Denmark). Conversely, in Austria, Luxembourg, and the United Kingdom, the coefficients are positive—indicating that increased renewable energy generation correlates with higher CO₂ emissions in the short run, with values of 0.55, 0.37, and 2.47, respectively. In Belgium, Ireland, and the Netherlands, the short-run impact of renewable energy on CO₂ emissions is not statistically significant. Regarding trade openness, no statistically significant effect is found in Belgium, Denmark, Finland, Germany, Ireland, Italy, or the United Kingdom. In Austria, France, and Luxembourg, increased trade openness raises CO₂ emissions, with coefficients of 0.53, 0.17, and 0.72, respectively. In contrast, in the Netherlands, a 1% increase in trade openness reduces CO₂ emissions by 0.33%.

Table 4 also presents the values of the error correction mechanism (ECM), which range from −0.11 in France to −1.12 in the Netherlands. All values are statistically significant at the 1% level, confirming the presence of a stable long-term relationship between CO₂ emissions, renewable energy, trade openness, and economic growth. The results of the diagnostic tests are summarized in

Table 5. Diagnostic test results.

Dependent Variable	Austria	Belgium	Denmark	Finland	France	Germany
ΔlnCO2t
JB	1.37 (0.50)	2.14 (0.34)	0.90 (0.64)	0.91 (0.63)	2.08 (0.35)	0.08 (0.96)
ARCH	0.49 (0.47)	2.74 (0.10)	0.94 (0.32)	0.40 (0.52)	0.19 (0.66)	2.26 (0.12)
BG	0.02 (0.97)	0.01 (0.98)	2.82 (0.08)	2.44 (0.11)	2.08 (0.16)	0.53 (0.60)
Dependent variable	Ireland	Italy	Luxembourg	The Netherlands	United Kingdon
ΔlnCO2t
JB	1.06 (0.59)	0.05 (0.97)	0.26 (0.88)	0.38 (0.83)	0.75 (0.68)
ARCH	1.44 (0.24)	1.08 (0.31)	3.10 (0.09)	0.71 (0.40)	0.08 (0.77)
BG	2.07 (0.15)	1.22 (0.32)	1.06 (0.36)	2.43 (0.12)	0.10 (0.91)

JB—Jarque–Bera test; ARCH—heteroskedasticity test F stat (p-value); BG—serial correlation LM test F stat (p-value). Source: Authors’ calculations.

. These indicate that the model residuals follow a normal distribution and that there is no evidence of heteroscedasticity or autocorrelation, thereby supporting the robustness and reliability of the ARDL model estimations.

5.4. The Robustness Check

The robustness check of the results is conducted using panel econometric methods. The analysis covers 12 EU countries. These countries were selected based on prior stationarity testing, which confirmed that none of the variables are integrated of order two, I(2)—a key prerequisite for panel ARDL estimation. The time span of the analysis covers the period 1980–2022, resulting in a balanced panel with a total of 516 observations. The empirical findings of the panel analysis are reported in

Table 6. Panel cointegration test result. Dependent variable: lnCO_2it. Selected model: ARDL (2, 0, 2, 0).

Test	Statistics	Probability	Residual Variance	HAC Variance
Kao	−1.31	0.0957	0.003	0.003
Long-run analysis
	Coefficient	Std. error	Statistics	Probability
lnGDPt	0.95	2.83	3.38	0.00
lnREt	−0.23	0.09	−2.41	0.02
lnTOt	−1.03	0.31	−3.28	0.00
Constant	27.47	4.61	5.95	0.00
Short-run analysis
ECM-1	−0.02	0.005	−3.00	0.00
ΔlnREit	−0.08	0.31	−2.44	0.01
ΔlnTOit	−0.05	0.03	−1.54	0.12

Source: Authors’ calculations.

.

First, [65]’s panel cointegration test confirmed the existence of a long-term relationship among the variables. Following this, the PMG-ARDL model was estimated to explore both short- and long-term dynamics. The results show that an increase in the share of renewable energy sources leads to a reduction in CO₂ emissions in both the short and long term. The corresponding coefficients are −0.08 (short run) and −0.23 (long run), respectively. These findings are consistent with those reported in previous studies using panel data methods [78,79,80]. In addition, economic growth is found to be associated with an increase in CO₂ emissions, which aligns with the conclusions of [9,81,82]. On the other hand, trade openness appears to have a mitigating effect on CO₂ emissions—contrary to the findings of [79].

Overall, the panel-based robustness check largely confirms the results obtained from individual country-level time series analyses, thereby strengthening the reliability of the main findings.

6. Concluding Remarks and Policy Implications

This study examined the effects of economic growth, renewable energy use, and trade openness on CO₂ emissions in the EU-15 countries over the 1980–2022 period. Given the EU’s commitment to carbon neutrality by 2050, the analysis contributes to understanding the region’s progress toward its environmental goals. Methodologically, this study employed the Autoregressive Distributed Lag (ARDL) technique to assess country-level relationships, with panel PMG-ARDL used for robustness checks. Due to the higher integration order of GDP in Greece, Portugal, and Spain, these countries were excluded from the ARDL analysis. Similarly, Sweden was excluded after no cointegration was found. Thus, the results are based on 11 EU countries.

Long-run analysis shows that in 8 out of 11 countries, GDP growth significantly increases CO₂ emissions, suggesting that economic growth remains carbon-intensive. These results were supported by panel-level estimates. Renewable energy was found to reduce emissions in seven countries, reinforcing its role in climate mitigation. However, in Germany, renewable energy use was associated with increased emissions, likely due to transitional inefficiencies or incomplete substitution. Trade openness exhibited mixed and mostly statistically insignificant effects on emissions in both the short and long run, highlighting the complexity of trade–environment dynamics. In the short run, GDP was not a significant driver of emissions in most cases. Renewable energy reduced emissions in five countries, while the effects of trade were largely negligible. These results imply that decoupling economic growth from CO₂ emissions has not yet been achieved in most EU-15 countries. Countries must therefore reexamine their economic structures and accelerate investment in clean technologies, energy efficiency, and circular economy strategies. No uniform pattern was found in the trade–emissions relationship. In Belgium, Denmark, and the Netherlands, trade openness correlated with reduced emissions, likely due to the export of cleaner products. In Austria, Finland, France, Germany, and Italy, trade had a neutral effect, suggesting decoupling. In Ireland, Luxembourg, and the UK, trade growth increased emissions, requiring targeted trade and energy policy reforms to promote cleaner imports and exports.

The EU’s CO₂ emissions declined by 2.2% in 2024, with renewables accounting for nearly 50% of electricity production—a positive trend driven by wind, solar, and hydropower. However, more action is required, especially in countries like Germany, where the long-run impact of renewable energy remains counterproductive. Countries should phase out coal and inefficient fossil fuel systems; improve energy system efficiency, especially in Germany through enhanced wind and solar deployment; make clean energy more affordable and accessible than fossil fuels; expand carbon pricing mechanisms to incentivize clean technology adoption; and ensure policy consistency through long-term regulatory frameworks to attract private investment in renewables.

It is important to note that short-run coefficients often contradict long-run estimates, which highlights the need for dynamic, adaptive policymaking. In the short term, investments in renewables may increase emissions due to transitional reliance on fossil fuels and infrastructure build-up. Over time, however, renewable energy significantly reduces emissions, once deployment is widespread and technology matures. Therefore, short-term anomalies should not deter renewable energy adoption. Policies should focus on long-term decarbonization, promote clean technology innovation and energy storage, and use transitional tools like temporary carbon taxes or clean energy subsidies.

In the context of achieving a sustainable path, the EU must continue to emphasize the EU ETS scheme in order to further reduce CO₂ emissions (since its introduction in 2005, emissions in heavy industry have decreased by around 40%). This can also be achieved by further expanding its scope, as was the case in 2024 when maritime transport was included. At the same time, carbon leakage should not undermine competitiveness and opportunities for sustainable growth in the EU-15. In this regard, custom measures such as the Carbon Border Adjustment Mechanism (CBAM), which was introduced in October 2023, are crucial for sustainability. The CBAM involves applying a carbon price to imported goods originating from the heavy industry. In this case as well, a proactive approach is necessary—both in terms of pricing (as projections show an exponential increase in the cost of importing one ton of CO₂) and in the scope of emissions covered.

To enhance the effectiveness of decarbonization efforts, several additional strategic directions should be considered. First, the social dimension of the energy transition must be prioritized through the implementation of Just Transition principles, ensuring support for workers and communities affected by the decline in fossil fuel-dependent industries. Second, stronger coordination across governance levels is essential to align local, national, and EU-wide strategies, with region-specific approaches that account for varying energy structures and development levels. Third, public spending and fiscal policy should be leveraged to accelerate decarbonization, particularly through green public procurement and the gradual elimination of fossil fuel subsidies. Moreover, digital technologies—including artificial intelligence and smart grids—should be integrated to improve energy efficiency and optimize consumption. The financial sector also plays a critical role through the development of green financial instruments and the incorporation of climate risks into investment decisions. Finally, establishing robust mechanisms for the monitoring, evaluation, and revision of climate policies, along with active participation in international climate frameworks—especially regarding technology transfer and carbon trade regulations—will be essential for ensuring the long-term success of the EU-15 climate agenda.

Despite its contributions, this study has several limitations: the time series length could be extended to enhance robustness; the geographical scope is limited to the EU-15, restricting generalizability; more formal econometric tests (e.g., multiple cointegration and unit root tests) could improve methodological rigor; testing the EKC directly could yield further insights; and the inclusion of variables like urbanization and financial development and testing alternative proxies for renewable energy would add depth and validate the findings.

The EU-15 countries face ambitious challenges, with economic growth and competitiveness at the core of their policy agenda. On 29 January 2025, the European Commission published a Communication titled “A Competitiveness Compass for the EU” to guide its strategic direction over the next five years. The Compass identifies three key priorities for enhancing EU competitiveness: closing the innovation gap, decarbonizing the economy, and reducing strategic dependencies. In this context, the analyzed countries should formulate targeted strategies to support the decarbonization process. These include the following: (i) advancing a Clean Industrial Deal to help reduce carbon emissions—particularly in energy-intensive industries—and support their transition to low-carbon technologies; (ii) presenting tailor-made action plans for vulnerable sectors such as chemicals, steel, and metals; and (iii) developing an Affordable Energy Action Plan aimed at lowering energy prices and reducing overall energy costs.