Exploring the Nexus between Greenhouse Emissions, Environmental Degradation and Green Energy in Europe: A Critique of the Environmental Kuznets Curve

Abstract

This study examines the intricate relationship between economic growth and European environmental degradation via the Environmental Kuznets Curve (EKC). Our results contest the traditional inverted U-shape model of the Environmental Kuznets Curve, indicating that the theory may not be consistently applicable across European countries. Utilizing CS-ARDL and MMQR modelling, we reveal substantial regional disparities. Western European nations demonstrate a typical Environmental Kuznets Curve (EKC) pattern in the short term, characterized by an initial increase in emissions alongside GDP development, followed by a subsequent fall. Conversely, Eastern and Balkan nations exhibit a U-shaped connection, described by an early decline in emissions followed by a subsequent increase as their development levels increase. The influence of renewable energy differs, as it decreases emissions in the short term in Western Europe. However, its long-term impacts are variable, especially when contrasted with its more pronounced effect on emissions in Eastern and Balkan countries. Furthermore, trade openness intensifies environmental degradation in the short-term across all regions, although its long-term impact diminishes, particularly concerning greenhouse gases (GHG). The relationship between renewable energy and trade openness is substantial for the short-term reduction of carbon dioxide emissions, but this effect declines with time. The results indicate that a uniform environmental policy throughout Europe may lack efficacy. Customized strategies to expedite the transition in Western Europe and more specific interventions in Eastern Europe are essential to harmonize economic progress with environmental sustainability. Future research should examine the determinants of the diminishing long-term effects of renewable energy and the interplay between trade and environmental policies.

1. Introduction

The Environmental Kuznets Curve hypothesis is a highly contested concept in environmental economics. It postulates that as a country’s GDP rises, its environmental degradation level rises, but declines once a certain income threshold is reached [1,2]. The concept gained considerable attention in the 1990s due to its implications for the correlation between economic growth and environmental sustainability. This was highlighted by the works of Grossman and Krueger [3,4], Panayotou [5], Shafik [6] and Selden and Song [7]. A correlation between GDP and local air and water pollutants is often observed, with pollution levels following a bell-shaped curve. This tendency suggests that since per capita income levels start low, per capita emissions or concentrations generally rise, albeit slower. Emissions or concentrations decrease once a specific income threshold is reached as income increases. It is important to note that the research is not definitively conclusive regarding global pollutants such as CO₂ [8].

This hypothesis forms the foundation of most environmental measures undertaken by governments. Given that the European Union is the global frontrunner in combating climate change and was the first significant economic bloc to establish a legally binding framework for all its member states (27 as of 2021) to fulfil the commitments of the Paris Agreement [9], it is crucial to maintain a focus on and provide backing for environmental policies grounded in consistent theories. By pursuing this approach, we may progress in attaining worldwide sustainability objectives. In this context, in 2019, the EU launched the European Green Deal, which aims to achieve climate neutrality as the primary objective for addressing environmental harm [10]. The EU has strengthened its efforts to achieve carbon neutrality by 2050. This was prompted by the decrease in greenhouse gas (GHG) emissions during the lockdowns in 2020 and 2021. As a result, the EU has introduced the “Fit for 55 package”, which aims to significantly reduce emissions through the EU European Trading System (ETS) mechanism. Precisely, a 40% reduction in emissions by 2030 is planned.

Nevertheless, upon closer scrutiny of the Environmental Kuznets Curve (EKC), various deficiencies become apparent, which require a thorough assessment. A major critique of the Environmental Kuznets Curve (EKC) is its tendency to oversimplify the complex relationship between economic growth and environmental outcomes [11]. The hypothesis assumes a universal, homogenous correlation between income and environmental quality, disregarding the various contextual elements that can impact this correlation. In addition, the EKC hypothesis has been criticized for its failure to explain the diversity of environmental problems, as different forms of pollution may have distinct associations with income levels [12]. Furthermore, the EKC theory has faced criticism based on its methodology, as several academics contend that the observed inverted U shape might be a statistical artefact rather than an accurate depiction of the underlying link between emissions and GDP levels [13]. Choosing suitable econometric methods and effectively addressing the possibility of omitted variable bias are important factors to consider when doing empirical research on the Environmental Kuznets Curve (EKC). Most studies employ first-generation data models such as fixed effects, random effects, Generalized Least Squares, FMOLS, DOLS, ARDL and quantile regression [14,15,16,17]. Even though several authors found that cross-section dependency is present, they have chosen to employ models that do not address this issue. Other authors have used CCEMG and AMG estimators to address cross-section dependency without considering endogeneity [18]. These aspects have not been systematically addressed in the current body of literature, which reveals a research gap that is addressed in the current research.

In addition, the EKC hypothesis has faced criticism regarding its normative implications. It implies that environmental degradation is an unavoidable result of economic growth and that the solution lies in reaching a specific income level rather than actively implementing environmental protection policies [12]. The hypothesis of the Environmental Kuznets Curve proposes a non-linear link between environmental degradation and income per capita, where deterioration first increases with income but then decreases after reaching a certain threshold. Nevertheless, this hypothesis has encountered numerous critiques over time. Carson [19] contends that the initial empirical support for the Environmental Kuznets Curve (EKC) is feeble and susceptible to variations in data and methodology. Research utilizing more robust data and advanced methods reveals diminished backing for the Environmental Kuznets Curve (EKC).

Furthermore, he emphasizes that studies on the Environmental Kuznets Curve (EKC) frequently demonstrate correlation rather than causation. Although there is a correlation between economic growth and a decrease in pollution, it is essential to note that this relationship may be attributed to improved regulation, technological advancements or more public awareness rather than income alone. Webber and Allen [20] argue that while the Environmental Kuznets Curve (EKC) may exist, the specific moment environmental improvement begins is undetermined and might take many decades. This indicates substantial harm to the ecosystem before reaching that stage.

To summarize, although the Environmental Kuznets Curve (EKC) offers a thought-provoking framework for understanding the connection between economic growth and environmental quality, it is essential to acknowledge its constraints and refrain from relying solely on it for making environmental policy decisions. Although the EKC theory has garnered much attention, it is necessary to critically evaluate its validity and applicability due to its limits and weaknesses. Addressing these methodological concerns is essential to accurately analyse the relationship between environmental degradation and income levels. Our research is focused on improving the robustness of previous empirical studies on the EKC by implementing rigorous econometric techniques and controlling for potential biases.

Therefore, this study makes two distinct contributions to the existing body of literature on EKC: firstly, by demonstrating that the Environmental Kuznets Curve (EKC) hypothesis is not valid in the long term when using suitable econometric models and, secondly, by offering a thorough examination of the connections between renewable energy, trade openness, carbon dioxide emissions and greenhouse emissions. We use a CS-ARDL model that includes the possible interaction effects between trade openness and renewables to understand these factors’ direct and indirect effects and how they might affect each other. Additionally, our findings offer important insights for policymakers seeking to design effective strategies for reducing carbon dioxide emissions and promoting sustainable development, considering that economic growth will continue to lead to environmental degradation in the long run. Another key feature of our research is that we employ a parsimonious model to avoid adding numerous independent variables in the econometric modelling that are not directly related to the studied phenomenon. This allows us to focus on the most relevant factors and produce more accurate results in the analysis. By using a parsimonious model, we can simplify the complexity of our research and make it easier to interpret the data. This approach also helps us avoid overfitting our model and ensures that our findings are robust and reliable.

This study’s main contribution is demonstrating that the Environmental Kuznets Curve (EKC) does not hold in the long run for European countries. Consequently, this finding raises doubts about the efficacy of present environmental regulations founded on this theory. This necessitates policymakers considering additional strategies for attaining sustainable development goals. Contrary to expectations, our study found that trade openness and renewable energy do not individually reduce environmental degradation over time. However, when these two factors interact, they have a beneficial combined impact on environmental sustainability. This suggests that an integrated approach, combining trade policies with renewable energy investments, may be more effective in achieving environmental goals. Policymakers should consider these interactions when designing strategies for sustainable development.

The paper is structured into five sections, as outlined below. Section 2 overviews the current research field and the latest developments. Section 3 extensively describes the approach employed, including the data collection process and the formal estimation model presentation. Section 4 presents an overview of the descriptive statistics, test results and estimates. It also includes a discussion of the findings. Section 5 provides a concise summary of the main findings. We also address any constraints in the research and suggest potential avenues for future investigation.

2. Literature Review

The Environmental Kuznets Curve (EKC) theory is used in environmental policy and sustainable development strategies [21], which might put the theory under scrutiny due to its fragile assumptions [22,23]. Stern [24] states that the EKC is an inverted U-shaped relationship between environmental degradation and income per capita. Environmental degradation initially increases and then reduces with growing income per capita [25,26], which underlines that the environmental quality worsens at the initial phases of economic development and improves afterwards [13].

The discussion regarding pollution, climate change and emissions mainly centres around carbon dioxide (CO₂) emissions. The EKC hypothesis is supported by research that also considers the relationship between economic growth and other factors such as globalization, population density, energy consumption, trade openness and CO₂ emissions. Most findings support the Environmental Kuznets Curve (EKC) theory (

Table 1. Summary of articles that employ quantitative methods and their results.

Reference	Environmental Degradation Proxy (Air)	Countries	Sample Period	Results
Grossman and Krueger [3,4]	SO2, Smoke, Heavy Particles	42 (Worldwide)	1977, 1982, 1988	EKC is confirmed except for Mexico
Panayotou [5]	SO2, NOx, Suspended Particles Matter (SPM)	68 (Worldwide)	1980s	EKC is confirmed for all countries
Selden and Song [7]	SO2, NOx, CO, SPM	30 (Worldwide)	1973–1975, 1979–1981, 1982–1984	EKC confirmed for all countries
Shafik [6]	SPM, CO2 and SO2	149 (Worldwide)	1960–1990	EKC confirmed for SPM and SO2, while for CO2, the results are mixed
Stern et al. [27]	SO2	Worldwide *	1990	EKC is not confirmed
Perman and Stern [28]	SO2, CO2	74 (Worldwide)	1960–1990	EKC is not confirmed
Galeotti et al. [8]	CO2	24 (OECD)	1960–2002	For 12 countries, the EKC stands, while for the other half, the EKC is not confirmed.
Apergis and Ozturk [25]	CO2	14 (Asia)	1990–2011	EKC is confirmed for all countries
Ulucak and Bilgili [29]	Ecological footprint (EF)	Worldwide	1961–2013	EKC is not confirmed
Destek et al. [30]	Ecological footprint (EF)	EU countries	1980–2013	EKC is confirmed for all countries
Sarkodie and Strezov [21]	CO2	China, India, Iran, Indonesia and South Africa	1982–2016	EKC is confirmed only for China and Indonesia
Gorus and Aslan [31]	CO2	Middle East and North African countries	1980–2013	EKC is partially confirmed
Koengkan et al. [32]	CO2	Latin America and Caribbean countries	1990–2016	EKC is not confirmed
Kostakis et al. [14]	CO2	Middle East and North African countries	1994–2014	EKC is confirmed for all countries
Kostakis et al. [15]	CO2	20 (EU)	2004–2018	EKC is partially confirmed
Bjørnskov [33]	CO2 and Total Greenhouse Emissions	155 (Worldwide)	1975–2015	EKC is not confirmed for democratic countries
Choong et al. [18]	CO2	76 countries	1971–2014	Confirmed in 16 out of 76 countries but does not fit all countries (CCEMG estimator) and confirmed in 24 out of 76 countries based on AMG estimator
Pata and Caglar [16]	CO2, Ecological footprint (EF)	China	1980–2016	EKC is not confirmed
Bimonte and Stabil [34]	Emission of Building Permits per capita	20 Italian regions	1980–2008	EKS is not confirmed
Isik et al. [17]	CO2	8 OECD countries (USA, Turkey, Australia, Canada, France, Sweeden, Netherlands, Portugal)	1962–2015	EKC is confirmed for 4 out of 8 OECD countries
Dogan and Inglesi-Lotz [35]	CO2	European countries	1980–2014	The EKC is not confirmed when industrial share is used but is confirmed when aggregate growth is used.
Pata [36]	CO2, Ecological footprint (EF)	USA	1980–2016	The EKC is confirmed

Source: Authors’ work. * The countries studied by Stern [19] considered data from the World Bank Development Report from 1992.

).

Economic growth has three main channels through which it impacts the environment [37]. Firstly, as economies expand, there is an increase in production and consumption, resulting in more resource extraction and waste generation. This can potentially harm the environment. Secondly, economic growth can cause a shift in the types of industries and products being produced and consumed. This shift may involve moving away from polluting industries towards cleaner sectors, which could benefit the environment. Lastly, economic growth often drives technological innovation, leading to more efficient production processes and less polluting technologies. This has the potential to reduce environmental damage. Grossman and Krueger [3,4] conducted analyses investigating the correlation between different pollutants, yielding inconclusive findings. While certain pollutants display an Environmental Kuznets Curve (EKC) pattern, others demonstrate a straightforward correlation with income or a more intricate relationship.

Moreover, it was observed that during the initial growth phases, environmental deterioration intensifies because of variables such as industry, urbanization and restricted environmental controls. As countries achieve higher levels of prosperity, there is a tendency for environmental quality to improve. This is primarily due to implementing more stringent rules, technological developments and increased public awareness, which is a more nuanced approach [5]. Nevertheless, this correlation is not inherent and relies on other circumstances, such as

• Policy choices playing a vital role in reducing environmental damage during the initial phases of development and expediting improvements as affluence increases;
• Technological progress potentially facilitating the separation of economic expansion from environmental degradation by permitting cleaner industrial processes and resource-efficient technology;
• Heightened environmental consciousness among the public potentially stimulating the need for more stringent legislation and eco-friendly goods and services.

As proven by Shafik [6], even in the case of deforestation, there is also a non-linear relationship with income per capita, thereby providing support for the Environmental Kuznets Curve (EKC) theory. Deforestation exhibits an initial positive correlation with wealth, but decreases when countries attain higher levels of affluence. While many indicators, such as access to safe drinking water, seem to increase as income rises, others, like faecal pollution, exhibit a more intricate correlation with income. The results regarding air pollution are inconclusive. Urban air pollution typically follows initial deterioration and subsequent improvement as wealth increases, indicating an Environmental Kuznets Curve (EKC) pattern. However, carbon dioxide emissions consistently rise as income increases, without any decline. Shafik’s analysis reveals that the correlation between economic progress and environmental conditions is intricate and variable based on the particular environmental measure under examination. Although certain signs in his model support the EKC theory, it is not generally applicable.

For example, Ulucak and Bilgili [29] reevaluated the Environmental Kuznets Curve (EKC) hypothesis by examining countries in three income groups using the Ecological Footprint as a proxy for environmental degradation. The authors’ findings verified the EKC hypothesis for all three categories: low, middle and high-income countries. This theory holds for countries in the Middle East and North Africa. Specifically, most indicate that energy consumption harms environmental quality, while globalization contributes to environmental degradation [31,32,35,36].

Bjørnskov [33] employed the Economic Freedom of the World (EFW) index to assess the Environmental Kuznets Curves for CO₂ and other greenhouse gases. The findings indicate that the shift towards reduced emissions happens more rapidly and at lower income thresholds in economically independent societies. Countries with higher wealth and greater economic freedom, such as the USA, Canada, Australia and those in Northern Europe, have already surpassed the CO₂ Kuznets curve. Increased economic development is expected to decrease emissions in most Western cultures, provided that their policies constantly emphasise economic freedom. Furthermore, Galeotti et al. [38] critically analysed past EKC studies, highlighting their failure to sufficiently consider the complex nature of this relationship and their reliance on possibly erroneous statistical methodologies. The authors refute the understanding of the oversimplified Environmental Kuznets Curve (EKC). Although certain countries provide some evidence favouring the Environmental Kuznets Curve (EKC), these findings suggest that the correlation between income and CO₂ emissions is inconsistent and may fluctuate over time. This implies that elements other than wealth, such as technical progress, policy modifications and economic structural changes, substantially impact environmental conditions. Therefore, more advanced econometric methods and a more profound comprehension of country-specific characteristics are needed to precisely evaluate the EKC hypothesis and its consequences for environmental policy.

Additional research has investigated the correlation between environmental deterioration and variables such as economic growth, the utilization of renewable energy sources, higher education attainment, the rigorousness of environmental policies and the level of female participation in parliamentary bodies. In their study, Kostakis et al. [15] showed that the Environmental Kuznets Curve (EKC) hypothesis applies to 20 European Union countries. This implies that as economies develop, environmental deterioration occurs non-linearly. The study also emphasizes the significance of stringent environmental legislation and the involvement of women in decision-making. The statement implies that sustainable development policies can mitigate economic distress while enhancing environmental quality. In a previous study, Koengkan et al. [32] provided evidence of a positive relationship between gender inequality, as measured by the difference in salary between genders and an increase in CO₂ emissions.

Economic growth, reduced greenhouse gas emissions and enhanced technologies are all facilitated by adopting renewable energy, which is essential for environmental prosperity. Renewable energy also mitigates ecological footprints and CO₂ emissions [39,40]. Moreover, Horobet et al. [41] investigated the relationship between economic growth and European emissions, indicating that economic growth can be achieved without causing environmental damage. In other research, Horobet et al. [42] conducted a 22-year study in 1996 to examine the connection between financial development and environmental degradation within the European Union. The research determined that specific institutions offer financial products and instruments that cause ecological damage to households and businesses. However, Fareed et al. [43] also discovered that financial inclusion in the Eurozone results in environmental degradation. Similarly, Ibrahim and Vo [44] investigated 27 selected industrialized countries operating from 1991 to 2014 and discovered that pollution is exacerbated by enhanced financial development [45]. At the same time, research shows that innovative financial products can successfully support technological innovation and the adoption of renewable energy, which further positively impacts the environment [46,47].

Furthermore, other research has demonstrated that renewable energy is the most effective method of decreasing environmental degradation [48,49,50]. However, some contend that the production and disposal of renewable energy infrastructure may have adverse environmental consequences due to the increased intensity of renewable energy [51]. Nevertheless, other research has demonstrated that renewable energy does not significantly affect long-term CO₂ per capita emissions [52]. Consequently, various findings suggest that, despite renewable energy’s potential negative environmental repercussions, its overall advantages in reducing environmental degradation are substantial and should not be disregarded.

On the other hand, the EKC theory is a dilemma in the scientific community, with both supporters and critiques [53]. The EKC theory is criticized by many authors who consider factors such as income distribution, income inequality and pollution characteristics [13,38,54]. Other authors criticized the econometric techniques, conditioning variables and endogenous bias regarding the econometric methods [55,56,57]. One major drawback regarding the econometric model is represented by the involvement of the nonlinear transformations of nonstationary regressors and neglected cross-sectional dependence [58]. Data heterogeneity among turning points in research on the EKC was attributed to variations in the study period and econometric methods used in model estimation [54]. Also, Usman and Jahanger [59] mentioned the heterogeneity of the data and how it impacted the validity of the EKC, as institutional quality and remittance inflows have heterogeneous impacts on environmental degradation. Abdulwakil and Azam [60] proved that the heterogeneity in income levels influences the validity of the EKC hypothesis by demonstrating a long-run relationship between environmental degradation and key variables in middle-income Sub-Saharan African nations.

There is a lack of robust statistical foundation on the EKC hypothesis, as many studies underline the data stationarity and the need for cointegration to support the EKC relationship [24,38]. Another issue is represented by the parametric models used to estimate the EKC, which are inadequate to catch the actual shape of the relationship, leading to potentially distorted inferences [61]. Even more, this might be considered a gap in the systematic empirical decompositions and explicit testing of the EKC hypothesis, which leads to misunderstanding the fundamental drivers of the observed nexus [55].

Altıntaş and Kassouri [62] mentioned that the EKC hypothesis in Europe is sensitive to the type of environmental degradation alternative depleted, with the ecological footprint indicator being a better option than CO₂ emissions. At the same time, Doganc and Inglesi-Lotz [35] underlined that the EKC hypothesis is valid for European countries when considering aggregate GDP growth, but not the industrial share in GDP. Destek et al. [63] showed that the ecological footprint revealed a U-shaped link with actual income in EU countries, with non-renewable energy fostering environmental degradation and renewable energy and trade openness shrinking it. Furthermore, Boubellouta and Kusch-Brandt [64] said that the EKC hypothesis includes recognizing e-waste generation in the EU28+2, which means that the e-waste generation grows with the GDP until a certain threshold is achieved, after which it declines despite supplementary economic growth. Neagu [65] revealed the economic complexity in EU countries, demonstrating a reversed U-shaped curve, with a growth in energy intensity leading to a 3.9% expansion in CO₂ emissions.

The EKC is not validated by certain countries, though it is valid in others and is contingent upon the variables and period employed. According to Jóźwik et al. [66], Poland is the only Central European country with a valid EKC. The other nine countries experienced increased CO₂ emissions due to their energy consumption. In another study conducted on EU countries by Sterpu et al. [67], it was demonstrated that the EKC could not generate any conclusive evidence, as indicated by the inverted U-shaped and inverted N-shaped forms. The EKC is valid in certain circumstances, such as in the research conducted by Armeanu et al. [68]; it is not valid for primary energy consumption and greenhouse gas emissions, but it is valid for the emissions of sulphur oxides and non-methane volatile organic compounds in EU-28 nations. Bölük and Mert [69] asserted that the EKC was invalidated for carbon emissions in 16 EU nations. This underscored the limitations of the EKC as a scientific method, as it is contingent upon the countries, variables, econometric model and selected time frame. This is due to the unpredictability of the computational factors employed. The EKC is under scrutiny not only in Europe but also globally. Al-Mulai et al. [70] have identified Vietnam as an example of a country where the EKC does not hold, as evidenced by their study demonstrating that pollution increases as per capita income increases.

Conversely, Işık et al. [71] studied 10 US states to determine whether the EKC is valid. They found that the EKC is valid for five states in the US: Ohio, New York, Florida, Illinois and Michigan. Lorente and Álvarez-Herranz [72] conducted an additional study that proved that the EKC thesis was supported in OECD countries. This study underscored that promoting renewable energy can reduce greenhouse gas emissions and air pollution per capita.

Even though most of the studies confirm the EKC hypothesis, recent studies have employed more advanced techniques [34] and suggest that the inverted U shape is not a robust data feature and that the observed patterns are sensitive to model specification and estimation methods. In conclusion, despite the widespread acceptance of the EKC hypothesis, there is mounting evidence that econometric misspecifications may impact the supporting empirical findings. Careful re-examination of the underlying data and modelling approaches is necessary to determine the true nature of the relationship between economic development and environmental quality.

3. Materials and Methods

This section presents the main data sources and explains the choice of variables in the model. In addition, it discusses the methodology utilized and the additional robustness tests conducted to assure the reliability and accuracy of the findings.

3.1. Data

The data employed in this paper covers the period from 2004 to 2021 for European countries, with a total number of 480 observations across 28 groups: Austria, Belgium, Bulgaria, Croatia, Cyprus, Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden. The data were retrieved from two sources, namely Eurostat and Our World in Data (

Table 2. Variable symbol, definition, unit of measure and data sources.

Type of Variable	Notation	Variable Description	Unit of Measure	Description	Source
Dependent	CO2	CO2 Emissions	Tonnes	Carbon dioxide (CO2) emissions from fossil fuels and industry. Land-use change is not included.	Our World in Data
Dependent	GHG	Greenhouse gas emissions	Tonnes	Per capita greenhouse gas emissions include the carbon dioxide, methane and nitrous oxide from all sources, including land-use changes. They are measured in tonnes of carbon dioxide equivalents over a 100-year timescale.	Our World in Data
Explanatory	GDP	Purchasing power-adjusted GDP per capita	PPS (current prices), index EU27_2020 = 100 and coefficient of variation	GDP per capita is calculated as the ratio of GDP to the average population in a specific year. Basic figures are expressed in purchasing power standards (PPS), representing a common currency that eliminates the differences in price levels between countries to allow meaningful volume comparisons of GDP.	Eurostat
Explanatory	REN	Share of renewable energy in gross final energy consumption	Percentage	The indicator measures the share of renewable energy consumption in gross final energy consumption according to the Renewable Energy Directive. The gross final energy consumption is the energy used by end-consumers (final energy consumption) plus grid losses and self-consumption of power plants.	Eurostat
Explanatory	TRD	Trade openness/ Trade as a share of GDP	Percentage	The sum of the exports and imports of goods and services, divided by gross domestic product, is expressed as a percentage.	Our World in Data

Source: Authors’ work. Note: Data from Eurostat was retrieved from https://ec.europa.eu/eurostat/web/main/data (accessed on 25 May 2024). Data related to air pollutants retrieved from Our World in Data [73].

). All the data has annual frequencies.

The principal dependent variable used is carbon dioxide (CO₂) emissions. Carbon dioxide has long been recognized as a central indicator of environmental degradation, as it plays a key role in the global carbon cycle and is directly linked to various human activities that contribute to the deterioration of our natural ecosystems. The growing concentration of CO₂ in the atmosphere, primarily driven by the burning of fossil fuels and deforestation, has become a widely accepted proxy for measuring the extent of environmental degradation [74,75].

A substantial body of research supports using CO₂ as a proxy for environmental degradation. Previous studies have examined the relationship between globalization and environmental degradation, with carbon dioxide emissions often serving as a representative variable for this complex phenomenon [75]. Carbon dioxide is a byproduct of various industrial processes, particularly the combustion of fossil fuels for power generation and transportation. The increasing levels of CO₂ in the atmosphere are directly linked to the rise in global temperatures, ocean acidification and other adverse environmental consequences [76]. As previously observed by other authors [55], choosing the dependent variable (e.g., environmental degradation proxy) represents the most critical step on which EKC relies. Environmental degradation is a multifaceted issue encompassing various interconnected factors, such as natural resource depletion, pollution and habitat destruction [77,78].

Choosing the right proxy to measure and monitor environmental degradation is essential for accurately assessing the impact of human activities on ecosystems. Different proxies may be more suitable for specific types of degradation, such as deforestation or water pollution, so careful consideration should be given to selecting the most appropriate one for the desired outcome. We have decided only to consider air pollutants as our proxy for environmental degradation, as they directly indicate the human impact on the atmosphere and can provide valuable insights into overall ecosystem health. By focusing on air pollutants, we can more effectively track changes over time and assess the effectiveness of mitigation efforts in reducing environmental harm [79]. Regulatory bodies like the United States Environmental Protection Agency and the World Health Organization have established standards and guidelines for air pollutant concentrations [80]. Monitoring the levels of dispersed air pollutants and implementing long-term strategies to reduce them is necessary to safeguard the health of both resident populations and travellers [81]. Air pollutants have been studied from both human and environmental health perspectives and their measurement and tracking in air, water and food, as well as their exposure through residential proximity to waste sites, can provide valuable insights into the state of the environment. The Clean Air Act (USA) and Ambient Air Quality Directive (EU) define six “criteria pollutants”, namely, carbon monoxide, nitrogen dioxide, ozone, sulphur dioxide, lead and particulate matter, each with well-characterized health effects [82]. Therefore, we have considered robustness tests to change the main dependent variable with greenhouse emissions, including all previously mentioned pollutants (please refer to Table 2).

Renewable energy inclusion in the Environmental Kuznets Curve (EKC) models is feasible and highly recommended, given its enormous impact on environmental quality and economic growth. As countries switch to cleaner energy sources, the link between economic development and environmental deterioration shifts. Renewable energy adoption has the potential to accelerate the EKC’s turning point, resulting in environmental improvements at lower income levels. Several studies have found that including renewable energy consumption in EKC models improves their explanatory power and provides a complete picture of the growth-pollution nexus. For example, Bilgili et al. [83] discovered that using renewable energy cuts CO₂ emissions significantly across all income levels.

Similarly, Dogan and Seker [84] found that renewable energy use improves environmental quality in European Union countries. By incorporating renewable energy variables, researchers can better understand the complex dynamics of technological advancement, energy policy and their effects on environmental outcomes. This method enables more accurate predictions and policy recommendations, making it a necessary component of modern EKC modelling.

Due to the European continent’s high level of economic integration and cross-border commerce, incorporating trade openness into Environmental Kuznets Curve (EKC) modelling is paramount when examining countries in the region. Trade openness significantly impacts the relationship between economic growth and environmental deterioration via various pathways, including composition, technique and scale effects. For European countries, particularly those in the European Union, the free movement of commodities and services intensifies these consequences. Studies have found that trade openness can have both beneficial and negative environmental impacts, depending on environmental restrictions, technological diffusion and the nature of traded goods. For example, Shahbaz et al. [85] discovered that trade openness lowers carbon emissions in developed European countries, while Dogan and Turkekul [86] reached the same conclusion for the United States. However, Özokcu and Özdemir [87] found that trade openness may raise emissions in certain instances. By including trade openness in EKC models for European countries, researchers can account for these complex interactions, providing a more accurate representation of the growth–pollution nexus and allowing policymakers to develop more effective environmental strategies that consider international trade.

3.2. Methodology

Given the nature of our data and our research objective, we employed panel data modelling in the following general form:

$${ED}_{it}={\beta }_0+{\beta }_1{Y}_{it}+{\beta }_2{Y}_{it}^2+{\beta }_3{X}_{it}+{\epsilon }_t$$

(1)

where ED is the environmental degradation indicator—CO₂ emissions and GHG emissions, alternatively—Y is the GDP per capita, X is the vector of additional explanatory variables—renewable energy and trade openness, β₀ is the constant term, β₁ and β₂ are the coefficients for GDP terms, β₃ is the vector of coefficients for additional variables and ε stands for the error term.

In the context of panel data analysis, accurately estimating the extent of cross-sectional dependence is important for further statistical analysis [88]. Tugcu [89] proposed a methodology that considers two primary factors when estimating parameters in panel models: a limited number of observations (with more cross-sections than observed periods) and the possibility of cross-section dependency (CSD). The limited number of observations, with more cross-sections than observed periods, can introduce biases in the parameter estimates [90]. Additionally, cross-section dependency needs to be accounted for, as it can lead to inappropriate inferences on the coefficients [88]. The role of panel data in describing and explaining change has been well-documented [91].

In order to identify if cross-section dependency is present, we have employed the following tests: the cross-section dependence (CD) test [92,93], the power-enhanced CD test [94], the CD star test [95] and the weighted CD test [94]. The null hypothesis of weak cross-sectional dependence in each test means no significant cross-sectional dependence is present in the data. In other words, the observations within each cross-section are assumed to be independent.

Subsequently, the decision to use the second-generation unit root test (CIPS) over first-generation tests is based on its improved properties and ability to account for the existence of CSD. Unlike first-generation tests, which may not adequately capture these effects, the CIPS test addresses this issue [96], providing more reliable and accurate results when CSD is present. Then, we checked for slope homogeneity by applying the Pesaran and Yamagata [97] and Blomquist and Westerlund [98] tests to ensure the consistency of coefficients across groups.

Furthermore, we examined cointegration employing Kao [99], Pedroni [100,101] and Westerlund [102] to assess if there is a long-term relationship between the variables. Additionally, we employed the Granger test adapted by Juodis et al. [103], which suits data with a small number of periods and heterogeneous coefficients. Moreover, this test allows for cross-sectional dependence and cross-sectional heteroskedasticity.

Given that most of the variables display high kurtosis, indicating leptokurtic distributions with positive or negative asymmetry, it is well-documented in the literature that there is a possibility of the error term following a non-normal distribution [104]. Estimation using the Least Squares approach may not be reliable in these situations, as it is only effective when the residuals adhere to the Gauss–Markov assumptions and behave predictably [105]. Autoregressive models are frequently employed for analysing stationary time series because they efficiently address the autocorrelation commonly found in data streams [106]. Thus, we have utilised a cross-sectionally augmented autoregressive distributed lag model (CS-ARDL).

The autoregressive distributed lag model is a widely recognized model used to analyse the dynamic interactions between variables. Nevertheless, this model may be susceptible to the distributional assumptions of the data. Non-normal distributions have been frequently observed in diverse domains such as health, education and social sciences, as reported by researchers [107]. The cross-sectionally augmented autoregressive distributed lag (CS-ARDL) model, proposed by Chudik et al. [108], addresses this shortcoming by including extra variables that account for cross-sectional dependence. This allows the model to handle non-normal distributions better. This model expands upon the conventional autoregressive distributed lag framework by including cross-sectional information, which can improve the study of time series that display both cross-sectional and temporal interdependence. The CS-ARDL model is designed to resist different distributional assumptions, making it an appropriate option for analysing non-normal distributions and cointegrated non-stationary data [109]. In Stata 18.00, we initially computed the short-run coefficients and determined the long-run coefficients [110]. We implemented two more models that included a control variable for trade openness and the interaction between trade openness and renewables. Interaction arises when the impact of one independent variable on the dependent variable is contingent upon the level of another independent variable [111]. By including these additional variables, we were able to further explore the potential moderating effect of trade openness on the relationship between renewables and environmental degradation.

To ensure the robustness of our findings, we have performed the following: firstly, we grouped the countries in our panel data considering Western and Northern countries and Eastern and Balkan states. This division of countries considers historical evolutions but also development levels, as Eastern and Balkan countries are the former communist nations that entered the European Union in the latter stages (after 2007), while Western countries are the older members of the European Union. Similar approaches were taken in other studies—see, for example, Žarković et al. [112], Ganic [113], Simionescu et al. [114] and Horobet et al. [115]. Then, we implemented the CS-ARDL model on these two groups of countries. Moreover, to complement and validate our findings, we then conducted a Method of Moments Quantile Regression (MMQR) [116,117] on the entire panel dataset. The MMQR technique offers the advantage of examining the relationships between our variables of interest and emissions at different points along the emission distribution [118], providing insights into how these relationships might vary for low, medium and high emitters. By employing these different methodological approaches, the group-specific CS-ARDL models and MMQR regression, we were able to cross-validate our results.

4. Results and Discussions

Table 3. Descriptive statistics.

Variable	Obs.	Mean	Std. Dev.	Min	Max	Skewness	Kurtosis
CO2	540	7.75	3.37	2.90	25.98	1.96	9.57
GHG	540	9.85	4.04	3.27	27.96	1.43	5.68
GDP	540	26,672.78	12,526.53	7300	86,800	1.84	7.93
REN	540	19.49	14.24	0.10	77.36	1.52	5.67
TRD	540	139.77	78.32	45.42	393.14	1.03	3.10

Source: Authors’ work. Note: The statistics presented are at the sample level before log transformation. Obs. stands for the number of observations; Std. Dev. stands for standard deviation.

presents the descriptive statistics for the variables used in the implemented models. The data used in this research is a balanced panel, allowing more accurate comparisons between variables over time. All the variables are transformed into their natural logarithms. This transformation helps to ensure that the data are normally distributed and improves the interpretability of the results.

The mean values for the environmental variables suggest relatively high levels of carbon dioxide and greenhouse gas, with CO₂ and GHG averaging 7.75 and 9.85 units, respectively. The high standard deviations for these variables further indicate a significant degree of variability, potentially signalling the diverse nature of emissions across the countries in the sample. The skewness and kurtosis values provide additional information about the distributions of these variables. The positive skewness values indicate a right-skewed distribution, with most observations clustered at the lower end of the range with a few high outliers. The kurtosis values exceed 3, suggesting a leptokurtic distribution with a sharper peak and longer tails compared to the normal distribution.

Furthermore, we have noticed specific instances when the amounts of carbon dioxide (CO₂) and greenhouse gases (GHG) exceed the standard deviation range, as depicted in Figure 1 and Figure 2. Considering this, we tested structural breakdowns using the approach proposed by Ditzen et al. [115], which detects multiple breaks occurring at unknown periods. The test statistic we obtained is 1.99, lower than the critical values at the 1%, 5% and 10% significance levels. Consequently, we cannot reject the null hypothesis that these levels have no breaks. Although the absence of a structural break is implied, the test identified two estimated breakpoints, specifically in 2006 and 2013. These breakpoints account for the outliers observed at the ends of the whiskers during those periods. At the aggregated level, it is observed that in 2004, the median value of carbon dioxide emissions per capita was 7 tonnes, with the highest amount recorded in Luxembourg at 25 tonnes.

In contrast to 2021, the level has experienced a substantial reduction, reaching a mere 13.2 tonnes, representing the greatest level in the sample. The overall trend throughout the years indicates a decline for all European countries. Before the implementation of the European Green Deal in 2018, carbon dioxide (CO₂) emissions ranged from 15.5 tonnes to 3.2 tonnes, with a median value of 6.4 tonnes. In 2019, there was a substantial reduction in the disparity between the levels of different countries. The introduction of the European Green Deal in 2018 appears to have benefitted Europe’s reduction of per capita carbon dioxide emissions. The most significant reduction in CO₂ emissions can be found in Central-Eastern European countries, with an average decrease from 7.4 tonnes in 2004 to 6.7 tonnes in 2021, excluding Poland and the Czech Republic. The leading cause of this evolution is that both nations have historically relied heavily on coal, one of the most carbon-intensive resources, to generate power [119,120].

Meanwhile, greenhouse emissions per capita seem not to have such a quick reaction to the environmental policies implemented (Figure 2). In 2021, the level of greenhouse emissions had decreased by 10 tons for our sample, but the overall median has remained at a steady level of around 8 tons. The country with the highest level is Ireland, with 15.1 tons. According to the latest data from the Environmental Protection Agency, Ireland has made significant efforts to reduce GHG emissions by reducing coal, oil and peat use and increasing renewables in the energy mix. Even though in terms of CO₂, Ireland has a low level, methane is a significant pollutant contributor (Figure 3). Another curious aspect is that Finland in 2021 recorded 13.5 metric tons of GHG emissions while, for carbon dioxide, only 6.8 metric tons.

Similarly, Norway finds itself in the same context. This further shows that the other gases (as defined in the Kyoto Protocol) can lead to significant environmental degradation. On the other hand, Eastern European and Baltic countries have the lowest levels of GHG, as in the case of CO₂.

The GDP exhibits a mean value of $26,673 with a high standard deviation of $12,527, suggesting considerable variation in economic output across the sample. This implies that most of the countries in the sample have lower GDP values compared to a few countries with a significantly higher GDP. The positive skewness and high kurtosis value for GDP indicate a right-skewed distribution with a few high-value outliers. The country with the highest GDP per capita is Luxembourg, while the countries with the lowest values are Bulgaria and Romania. The share of renewable energy in gross final energy consumption has a mean of 19.49% and a standard deviation of 14.24%, suggesting a wide range of renewable energy utilization across the observations. The share of renewable energy in gross final energy consumption has a mean of 19.49% and a standard deviation of 14.24%, suggesting a wide range of renewable energy utilization across the observations. The variability is explained by countries having different levels of natural endowments, investment in renewable energy infrastructure and various policies promoting its use, leading to disparities in renewable energy consumption.

Related to the income level, countries vary significantly, with Bulgaria having the lowest GDP per capita at only 18,500 dollars and, at the other end with the highest-level, Luxembourg having 86,800 dollars (Figure 4). Data shows that even though wealthy countries have high economic growth, such as is the case of Luxembourg, Ireland and Norway, this does not lead to a decrease in environmental degradation in the long run. At the same time, the countries with the lowest GDP per capita also have low emission levels.

Countries with low GDPs per capita from Eastern and Western Europe have the lowest emissions per capita. This could be due to various factors, such as the lower levels of industrialization, smaller populations and less reliance on fossil fuels. We also observed that even though Nordic countries have high shares of renewable energy consumption (Norway with 74%, Sweden with 62.7% and Finland with 42.9%) (Figure 5), this is not significantly contributing to lower emissions.

Therefore, our study validates that the inconsistencies that can be observed relatively easily are also confirmed by the statistical analysis, providing further evidence of their existence. This strengthens the validity of our findings and underscores the importance of addressing these inconsistencies.

The cross-sectional dependence tests (

Table 4. Cross-section dependency tests and second-generation unit root test (CIPS).

Variable	Cross-Sectional Dependency Tests				Panel Unit Root (CIPS)
	CD	CDw	CDw+	CD*	Level	First Difference
lgCO2	57.98 *	−2.63 *	1346.79 *	−0.26	−2.069	−4.185 *
lgGHG	44.54 *	−2.65 *	1214.19 *	−0.21	−1.974	−4.321 *
lgGDP	72.6 *	−1.22	1599.02 *	1.55	−1.837	−3.515 *
lgGDPsq	72.64 *	−1.22	1599.57 *	1.61	−1.821	−3.502 *
lgREN	81.84 *	−2.28	1704.64 *	0.32	−2.398 *	−3.862 *
lgTRD	59.86 *	−0.47	1252.09 *	−0.92	−1.466	−2.972 *

Source: Authors’ work. Note: * denotes significance level at 10%. Cross-section dependency test results are in the following order: CD—Pesaran (2014, 2020) [92,93]; CDw and CDw+—Juodis and Reese (2022) [94]; CD*—Pesaran and Xie (2021) [95]. CIPS represents the second-generation unit test introduced by Pesaran (2007) [96], which was implemented at the level and first difference of the variables.

) yield varying results depending on the approaches used. Specifically, in the Pesaran CD-test (CD) context, all the variables exhibit CSD. The weighted CD test (CDw) and the CD star (CD*) provide comparable results, indicating that there is evidence of strong cross-dependency in the case of carbon dioxide and greenhouse emissions. Conversely, the null hypothesis is not rejected by any of the variables in the panel in the case of the power-enhanced CD test (CDw+). Considering the combined findings, we concluded that there is significant evidence of cross-sectional interdependence between the variables. Given the presence of CSD, we employed second-generation unit root tests and the Pesaran panel unit root test (CIPS), whose results indicate that all variables, except renewables, exhibit homogeneous non-stationary levels. However, all the mentioned variables are stationary in their first differences. In order to comfort this aspect, the estimations were run on the first difference.

The two slope heterogeneity tests [96,97] indicate that the slope coefficients are not homogeneous (

Table 5. Slope heterogeneity.

Test	Delta	p-Value
Blomquist and Westerlund (2013) [98]	8.13 *	0.00
Pesaran and Yamagata (2008) [97]	13.79 *	0.00

Source: Authors’ work. Note: * denotes significance level at 10%.

). These results suggest significant variation in slope coefficients across different countries. Therefore, we can conclude that the relationship between the independent and dependent variables does not remain consistent across all the observations. After careful consideration, we have determined that it is suitable to utilize the CS-ARDL method. This method is an ARDL version of DCCE incorporating individual estimations with lagged dependent variables and cross-section averages. Additionally, it enables the estimate of mean group values when the slope coefficients exhibit heterogeneity [101], which aligns with prior research [106,121].

The Kao ADF, Westerlund variance and Pedroni-modified Phillips–Perron tests (

Table 6. Cointegration test.

Cointegration Test	Statistic	p-Value
Augmented Dickey–Fuller	−1.22	0.11
Variance ratio	1.89 **	0.03
Modified Phillips–Perron	7.27 ***	0.00

Source: Authors’ work. Note: ** denotes significance level at 5%; *** at 1%.

) yield inconsistent findings. Although the Westerlund variance and the Pedroni test indicate that the panels are cointegrated, the Kao ADF test does not offer significant evidence of cointegration. Typically, cointegration tests reveal a long-term relationship between the variables of interest. Based on these data, we concluded that the variables are likely to move together over time, which supports the idea that a stable relationship exists between them. Additionally, we have checked for the multicollinearity of the data used in the estimations using the Variance Inflation Factor. We have found multicollinearity for “lgGDP” and “lgGDPsq”, which is expected, while for the remaining variables, we have not identified any correlation with the other variables in the model.

The results presented in

Table 7. Granger non-causality test (Juodis et al., 2021 [103]).

	lgCO2	lgGHG	lgGDP	lgGDPsq	lgREN	lgTRD
lgCO2	-	-	50.86 ***	47.23 ***	1.76	19.93 ***
lgGHG	-	-	34.85 ***	34.83 ***	1.27	26.2 ***
lgGDP	4.79 **	7.58 ***	-	-	0.65	7.67 ***
lgGDPsq	4.54 **	6.81 ***	-	-	0.18	7.03 ***
lgREN	3.97 **	87.37 ***	0.65	1.44	-	6.38 **
lgTRD	16.90 ***	36.65 ***	25.11 ***	25.17 ***	14.25 *	-

Source: Authors’ work. Note: The variable on the column does not have a Granger cause variable on the line. This test performs the half-panel jackknife (HPJ) Wald-type test for Granger noncausality. * denotes the significance level at 10%; ** at 5%; *** at 1%.

demonstrate that the connection between GDP and environmental degradation is mutually dependent. According to Hepburn et al. [122], economic growth stimulates a greater need for natural resources, which can have detrimental effects such as environmental degradation, habitat destruction, pollution and climate change. Environmental deterioration impacts the GDP through resource scarcity, health consequences and economic hazards. The detrimental effects of pollution and degradation can have a negative impact on public health, resulting in escalated healthcare expenses and reduced productivity in the workforce. Similarly, climate change substantially threatens economies, necessitating expensive recovery measures and posing long-term economic difficulties. Additional research has also emphasized the adverse effects of environmental degradation on economic growth. This includes the detrimental impact on both economic growth and biodiversity when environmental deterioration hampers local economic progress [123]. Moreover, the reciprocal relationship between environmental degradation and health outcomes in countries such as Pakistan highlights the interdependence of economic growth, environmental degradation and public health [124,125].

Renewable energy sources are causally linked to the degradation of the environment, as well as to trade openness. Regarding environmental deterioration, our findings indicate a one-way relationship, which differs from the findings of Sharif et al. [126], who identified a two-way causal connection. Our findings support the conclusions of Jebli et al. [127] about the relationship between renewable energy and trade openness. Specifically, our results indicate that countries with more open trade policies have better access to renewable energy technology and resources.

The empirical estimations in our baseline models (D1 and D3) and extended models (D2 and D4) in

Table 8. Estimation results.

Dependent Variable	CO2				GHG
Model Identifier	D1		D2		D3		D4
	Coeff.	P > |z|	Coeff.	P > |z|	Coeff.	P > |z|	Coeff.	P > |z|
Short-run estimations
LD.lggdp	31.738 **	0.04	35.716 *	0.09	24.244 **	0.05	28.019 **	0.02
LD.lggdpsq	−1.518 **	0.04	−1.743 *	0.06	−1.148 *	0.06	−1.337 **	0.03
LD.lgren	−0.012 ***	0.01	−0.433 ***	0.09	−0.167 **	0.03	−0.189 **	0.03
LD.lgtrd	-	-	0.637 *	0.01	-	-	0.195 *	0.07
LD.intvar1	-	-	−0.004 *	0.07	-	-	−0.003 ***	0.00
Long-run estimations
lr_lggdp	30.738 **	0.04	34.716 *	0.10	23.244 **	0.06	27.019 **	0.03
lr_lggdpsq	0.048 ***	0.00	0.049 ***	0.00	0.049 ***	0.00	0.052 ***	0.00
lr_lgren	0.000	0.84	0.005	0.17	0.000	0.98	−0.040 **	0.02
lr_lgtrd	-	-	0.001	0.83	-	-	−0.008	0.23
lr_intvar1	-	-	−0.001 ***	0.00	-	-	0.000	0.55
R-squared	69.00%		44.00%		67.00%		47.00%
Turning point in the short run	34,774		28,234		38,444		35,477

Source: Authors’ work. Note: * denotes significance level at 10%; ** at 5%; *** at 1%.

indicate that environmental deterioration increases with economic growth up to a certain threshold in the near term. This finding aligns with the current body of research [128,129], indicating that during the initial phases of economic growth, there is typically a rise in environmental degradation and depletion of resources. Nevertheless, the precise nature of the long-term link between these variables remains uncertain.

Regarding the Environmental Kuznets Curve, our findings indicate that the hypothesis holds true in the short term. However, we are unable to validate it in the long term. This aligns with the results obtained by A’yun [130] in the instance of India, which used a basic ARDL model without accounting for cross-section dependency. Due to technological advancements, globalisation, policy interventions and feedback loops, the correlation between economic growth and environmental sustainability may not exhibit a linear trajectory over time. Technological advancement is essential for maintaining environmental sustainability [131] and countries can enhance their ability to address environmental deterioration by investing in innovative technology. Globalization significantly influences sustainability issues worldwide [132] and the Environmental Kuznets Curve may not fully capture the complex dynamics of long-term environmental degradation.

Government and international organization policies can impact the course of environmental deterioration, resulting in outcomes that may not match the forecasts of the Environmental Kuznets Curve. Feedback loops can worsen environmental deterioration as time passes, leading to tipping points when environmental degradation speeds up, independent of economic levels [133].

Furthermore, the findings suggest that using renewable energy reduces adverse environmental effects. However, this correlation does not hold for long-term projections. Most research affirms that renewable energy sources, such as solar panels, wind turbines and biofuels, can alleviate environmental deterioration in the foreseeable future [30,34,38,39,134]. However, our findings indicate that, over time, they have no substantial impact on environmental sustainability. One possible explanation is that these activities have the capacity to lead to land degradation, biodiversity loss and water contamination. Hence, it is imperative to consider the entire life cycle of energy generation, encompassing the emissions linked to the manufacturing, transportation and installation of renewable energy systems. As the demand for renewable energy grows, there is a proportional rise in the pressure on natural resources needed to advance new technologies. Technological waste can arise from the limited operational lifespan of renewable energy sources, requiring their removal or recycling. Thus, using renewable energy sources might impose further pressure on ecosystems, resulting in problems such as deforestation, soil deterioration and diminished biodiversity [135]. Large-scale renewable energy projects like wind and solar farms can disrupt local species and ecosystems, harming biodiversity [136]. As attention shifts away from fossil fuels, renewable energy sources may cause local environmental deterioration in regions rich in renewable resources without addressing the global climate emergency. Our results suggest that renewable energy sources reduce emissions and immediate environmental harm, but a complete study of their long-term effects reveals subtleties and potential drawbacks.

Lastly, when examining models D2 and D4, it becomes evident that increased trade openness is strongly associated with a notable rise in environmental degradation. Multiple studies have highlighted the adverse effects of trade liberalization on environmental conditions. Chintrakarn and Millimet [137] discovered proof of trade’s environmental repercussions concerning carbon and nitrous emissions. In the same manner, Managi et al. [138] acknowledged the difficulties in accurately assessing the total influence of trade openness on the environment. Numerous arguments suggest that trade openness can contribute to environmental degradation through various processes. For instance, increased production and consumption might strain natural resources and ecosystems, leading to pollution, deforestation and resource scarcity [139]. According to Leal et al. [140], some countries specialize in industries with lax environmental regulations or harmful activities, which causes higher pollution levels and degradation. This situation is worsened by regulatory arbitrage, which can incentivize enterprises to move their output to countries with less stringent environmental restrictions, resulting in environmental degradation in regions with weaker regulations.

The interaction term “intvar1” in model D2 offers insights into the joint effect of trade openness and renewables on carbon dioxide emissions. The statistically significant coefficient indicates that a 1% increase in the interaction term, which represents a rise in trade openness, renewables, or a mix of both, would result in a 0.4% reduction in carbon dioxide in the short term. This suggests that the relationship between trade openness and renewables helps reduce adverse environmental effects. The long-term reduction in CO₂’s contribution to environmental degradation is just 0.1%, resulting in significantly less significant mitigation effects. This implies that although the influence of trade liberalization and renewable energy sources in decreasing carbon dioxide emissions is notable in the immediate future, it can gradually decrease effectiveness over time.

For model D4, the interaction term, in the short run, suggests that as trade openness and renewables increase, there is a negative impact on greenhouse gas emissions. However, the interaction effect is not statistically significant in the long term. This implies that the combined effect of trade openness and renewables on greenhouse gas emissions is not statistically significant over a longer time horizon. Various factors may be at play in the long term that could influence the relationship between trade openness, renewables and greenhouse gas emissions, leading to a lack of statistical significance in the interaction term. Possible reasons why the combined effect of trade openness and renewables on greenhouse gas emissions may not be statistically significant in the long run could include but are not limited to changes in technology, shifts in global energy markets, evolving environmental policies [141] and other external factors that can influence the relationship between these variables over time. Overall, the findings suggest that while trade openness and renewables may significantly affect greenhouse gas emissions in the short term, this relationship may not hold in the long term.

Table 9. CS-ARDL model estimations result on Country Groups.

Group 1
Dependent Variable	CO2				GHG
Model Identifier	D1		D2		D3		D4
	Coefficient	P > |z|	Coefficient	P > |z|	Coefficient	P > |z|	Coefficient	P > |z|
Short run estimations
LD.lgdp	21.79 **	0.02	19.15 **	0.02	19.33 ***	0.05	31.34 ***	0.00
LD.lgdpsq	−1.04 **	0.02	−0.90 **	0.02	−0.92 **	0.06	−1.49 ***	0.00
LD.lren	−0.03 ***	0.00	−0.28 *	0.06	−0.02 **	0.03	−0.16	0.31
LD.ltrd	-	-	0.07	0.38	-	-	0.04	0.70
LD.intvar1	-	-	0.00	0.00	-	-	0.00	0.99
Long run estimations
lr_lgdp	20.79 **	0.03	18.15 **	0.02	18.33 *	0.06	30.34 ***	0.00
lr_lgdpsq	0.05 *	0.00	0.00	0.31	0.05 ***	0.00	0.01 ***	0.00
lr_lren	0.00 **	0.09	0.05 ***	0.00	0.00	0.98	0.05 ***	0.00
lr_ltrd	-	-	0.02 ***	0.01	-	-	0.01 ***	0.00
lr_intvar1	-	-	−0.02	0.20	-	-	−0.01 ***	0.00
R-squared	0.56		0.55		0.56		0.57
Turning point in short run	36,476		42,154		36,117		35,750
Group 2
Dependent variable	CO2				GHG
Model identifier	D5		D6		D7		D8
Short Run estimations	Coefficient	P > |z|	Coefficient	P > |z|	Coefficient	P > |z|	Coefficient	P > |z|
LD.lgdp	−6.12 ***	0.00	−8.34 **	0.02	−7.14 ***	0.01	−10.77 ***	0.01
LD.lgdpsq	0.34 ***	0.00	0.44 **	0.01	0.39 ***	0.00	0.57 ***	0.01
LD.lren	−0.35 ***	0.00	−0.33 ***	0.00	−0.19 **	0.02	−0.20 **	0.04
LD.ltrd	-	-	0.24 ***	0.00	-	-	0.24 ***	0.00
LD.intvar1	-	-	−0.01 **	0.05	-	-	−0.01 ***	0.01
Long Run estimations
lr_lgdp	−7.12 ***	0.00	−9.34 ***	0.01	−8.14 ***	0.00	−11.77 ***	0.00
lr_lgdpsq	0.07 ***	0.00	0.05 ***	0.00	0.05 ***	0.00	0.05 ***	0.00
lr_lren	0.13	0.61	0.01	0.75	−0.03 **	0.01	−0.05 *	0.08
lr_ltrd	-	-	0.01 **	0.04	-	-	0.03 **	0.02
lr_intvar1	-	-	0.00	0.16	-	-	0.00	0.39
R-squared	0.48		0.49		0.61		0.58

Source: Authors’ work. Note: Group 1 countries consist of Austria, Belgium, Germany, Denmark, Spain, Finland, France, Ireland, Italy, Luxembourg, the Netherlands, Norway, Malta, Portugal, Sweden and Greece. Group 2 countries consist of Bulgaria, Cyprus, Croatia, Czech Republic, Hungary, Lithuania, Latvia, Poland, Romania, Slovenia and Slovakia. * denotes significance level at 10%; ** at 5%; *** at 1%.

presents the results obtained after the estimation of CS-ARDL on two country groups, namely, Group 1, which includes Western and Northern European countries, and Group 2, which comprises Eastern and Balkan countries.

Group 1 countries confirm the Environmental Kuznets Curve (EKC) hypothesis in the short run, where emissions initially increase with GDP growth, but eventually decrease at higher levels of economic development. The turning points for this relationship in Group 1 range from approximately EUR 35,750 to 42,154. In contrast to this, countries in Group 2 display U-shape relationships between the GDP and emissions in both the short run and long run, which means that initial economic growth is associated with low emissions but, as development progresses, emissions increase. The impact of renewable energy on CO₂ and GHG emissions also differs between the two groups. For Group 1 countries, renewable energy shows a tendency to reduce emissions in the short run, but its long-run effects are mixed and sometimes positive. On the other hand, for Eastern and Balkan states, it is observed that renewable energy leads to a decrease in emissions in the short run. This suggests that investments in renewable energy might be more effective in reducing emissions in ex-communist countries. Trade openness generally increases environmental degradation in both groups, indicating that increased trade tends to lead to higher emissions. However, the effect is more pronounced and statistically significant in the short run for Eastern and Balkan states.

The interaction between renewable energy and trade openness denotes varying effects across the two groups. In Western countries, this interaction has minimal short-run impact, but demonstrates a potential moderating effect in the long run, particularly for greenhouse emissions. Meanwhile, for Group 2 countries, the interaction term has a small but significant negative effect in the short run, suggesting a more immediate interplay between renewable energy and trade policies in these economies.

The contrasting trends identified between Group 1 and Group 2 countries regarding economic growth, environmental degradation, renewable energy and trade implications indicate that uniform policies in European countries may prove ineffective. For Group 1 countries, strategies may prioritize accelerating the transition to the declining segment of the Environmental Kuznets Curve, whereas, for ex-communist countries, initiatives may need to confront the issues presented by the U-shaped correlation between economic growth and emissions. The impact of renewable energy in mitigating emissions in ex-communist countries may guide strategic investments and regulatory incentives in these regions.

The MMQR estimations (see

Table 10. MMQR model estimation results.

Dependent Variable	CO2				GHG
Model Identifier	D1		D2		D3		D4
	Coefficient	P > |z|	Coefficient	P > |z|	Coefficient	P > |z|	Coefficient	P > |z|
Location
lgdp	3.66 ***	0.00	4.03 ***	0.00	4.14 ***	0.00	4.10 ***	0.00
lgdpsq	0.20 ***	0.00	0.22 ***	0.00	0.23 ***	0.00	0.23 ***	0.00
lren	−0.15 ***	0.00	−0.97 ***	0.00	−0.04 ***	0.01	−0.38 *	0.09
ltrd	-	-	−0.37 ***	0.00			−0.13	0.28
intvar1	-	-	0.17 ***	0.00			0.07	0.13
Scale
lgdp	2.06 ***	0.00	2.44	0.00	2.42 ***	0.00	2.61 ***	0.00
lgdpsq	−0.10 ***	0.00	−0.12	0.00	−0.12 ***	0.00	−0.13 ***	0.00
lren	−0.01	0.24	0.50	0.00	0.00	0.77	0.48 ***	0.00
ltrd	-	-	0.34	0.00			0.26 ***	0.00
intvar1	-	-	−0.09	0.01			−0.09 ***	0.00
Quantile 0.25
lgdp	5.56 ***	0.00	6.34 ***	0.00	6.50 ***	0.00	6.32 ***	0.00
lgdpsq	0.29 ***	0.00	0.33 ***	0.00	0.34 ***	0.00	0.33 ***	0.00
lren	−0.14 ***	0.00	−1.44 ***	0.00	−0.04 **	0.03	−0.79 ***	0.00
ltrd	-	-	−0.69 ***	0.00			−0.36 ***	0.01
intvar1	-	-	0.25 ***	0.00			0.15 ***	0.01
Quantile 0.50
lgdp	3.61 ***	0.00	4.23 ***	0.00	4.06 ***	0.00	4.23 ***	0.00
lgdpsq	0.20 ***	0.00	0.23 ***	0.00	0.22 ***	0.00	0.23 ***	0.00
lren	−0.15 ***	0.00	−1.01 ***	0.00	−0.04 ***	0.01	−0.41 *	0.08
ltrd	-	-	−0.40 ***	0.00			−0.15	0.23
intvar1	-	-	0.17 ***	0.00			0.07	0.11
Quantile 0.75
lgdp	1.98 **	0.02	1.98	0.13	2.31	0.00	2.02 *	0.06
lgdpsq	0.12 ***	0.00	0.12 *	0.07	0.14	0.00	0.12 **	0.02
lren	−0.16 ***	0.00	−0.55 **	0.04	−0.04	0.01	0.00	1.00
ltrd	-	-	−0.09	0.53			0.08	0.59
intvar1	-	-	0.09 *	0.10			0.00	0.95

Note: * denotes significance level at 10%; ** at 5%; *** at 1%.

) consistently show a positive relationship between economic growth and environmental degradation across all the quantiles and model specifications. This relationship is statistically significant in most cases, suggesting that their emissions tend to increase as economies grow. Renewable energy generally denotes a negative relationship with emissions, particularly in the lower and median quantiles. This suggests that the increased use of renewable energy is associated with lower carbon dioxide and greenhouse emissions, especially for countries or regions with lower to moderate emission levels. However, the effect seems to weaken or become insignificant at the 75th percentile, implying that renewable energy might be less effective in reducing emissions for the highest emitters.

Trade openness, included in models D2 and D4, shows a negative relationship with emissions in the location effects and lower quantiles. This suggests that increased trade openness is generally associated with lower emissions, possibly due to improved efficiency or technology transfer. However, the effect becomes insignificant at higher quantiles, indicating that the benefits of trade openness may diminish for the highest emitters.

The interaction term (intvar1) shows a positive coefficient in the location effects and lower quantiles. This indicates that the combination of increased renewable energy use and trade openness might have a complex effect on emissions, potentially offsetting some of the individual negative effects of these factors.

The scale effects reveal interesting patterns in the variability of emissions, confirming the Kuznets Curve, but only in the case of the variability of emissions. Renewable energy and trade openness tend to increase emission variability, suggesting that while these factors may reduce emissions on average, they might lead to more diverse outcomes across different contexts.

Across the quantiles, the effects of the variables tend to be the strongest and most significant at the 25th and 50th percentiles, with weaker or insignificant effects at the 0.75 quantiles. This pattern suggests that the variables considered have the most consistent impact on low to moderate emitters, while additional or different factors might influence high emitters (such as Western and Nordic countries). This remains an interesting avenue for further research.

5. Conclusions

Our research has yielded several key findings that question existing assumptions and shed light on the complex relationship between economic growth and environmental degradation by examining Europe’s Environmental Kuznets Curve (EKC). Our findings imply that the Environmental Kuznets Curve theory’s simple inverted U-shape model of the relationship between economic growth and environmental degradation may not be accurate for European countries, or at least not all. We also provide a more nuanced understanding of this complex relationship by incorporating various control variables and robustness checks. Moreover, by applying CS-ARDL and MMQR methodologies, we obtained richer insights into how different groups of countries in the region experience this relationship.

When all European countries are considered, our study provides evidence for a mutually dependent relationship between economic growth and environmental degradation, with the first causing an increased demand for natural resources, leading to environmental degradation. Secondly, the findings reveal that environmental degradation increases with economic growth up to a certain threshold in the near term, but the long-term link between these variables remains uncertain. The EKC hypothesis holds true in the short term but is not validated in the long term at the European level. Thirdly, increased trade openness is strongly associated with a notable rise in environmental degradation, as it can contribute to pollution, deforestation and resource scarcity.

Furthermore, we show that the joint effect of trade openness and renewables increases in the short run, positively impacting the environment. However, the combined effect of these factors is validated in the long term only for carbon dioxide. This implies that although the influence of trade openness and renewable energy sources in decreasing environmental degradation emissions is notable in the immediate future, it can gradually decrease effectiveness over time and even become insignificant. Further research could explore the reasons behind this decrease in effectiveness over the long run and identify ways to enhance the sustainability of this interaction. In addition, examining potential policy changes or technological advancements that could bolster the positive effects of trade openness and renewables on reducing carbon dioxide emissions would be valuable.

The detailed analysis of sub-panels of European countries revealed highly challenging findings for policymakers, as there are distinct dynamics in Western European countries compared to Eastern and Balkan states. Western European countries show a classic EKC pattern in the short run, where emissions rise with GDP growth and eventually decline at higher income levels, with a turning point for this relationship between $35,750 and $42,154, depending on the model used. However, in the case of Eastern and Balkan states, the relationship takes a U-shape, where emissions initially decrease with economic growth but increase again as development progresses. This divergence highlights that the EKC hypothesis is only partially valid across Europe, with significant regional variations.

Moreover, renewable energy has different impacts on emissions, depending on the group of countries. In Western Europe, renewable energy tends to reduce emissions in the short term, but its long-term effects are mixed. In contrast, for Eastern and Balkan countries, renewable energy shows a stronger and more consistent effect in reducing short- and long-term emissions, suggesting that investments in renewables may be more effective in ex-communist countries. Additionally, the role of trade openness in contributing to environmental degradation is noteworthy across both groups of countries. Trade openness increases emissions in the short run for both Western and Eastern European countries, though its effects are more pronounced in Eastern and Balkan states. Over the long run, trade’s impact on emissions tends to moderate but remains significant for GHG. Also meaningful is the interaction between renewable energy and trade openness within the European continent. In Western countries, this interaction has a minimal short-run impact but a potential moderating effect in the long run, particularly for GHG emissions. In contrast, the interaction term indicates a more immediate interplay between trade policies and renewable energy adoption for Eastern and Balkan countries.

Policymaking is greatly affected by these findings. Thus, governments cannot solve environmental problems with economic growth alone. Instead, strict environmental rules must be enacted early in the economic growth process. Our sustainability study recommends energy efficiency, circular economy principles, biodiversity conservation and renewable energy adoption. The discovery that renewable energy sources may not significantly influence environmental sustainability is especially troubling. It implies that the move to renewable energy, while necessary, is insufficient. Policies should prioritize comprehensive approaches to sustainability, such as energy efficiency, circular economy principles, biodiversity conservation and renewable energy adoption. From this perspective, the NextGenerationEU plan of the European Union, which is already under implementation, has a subcategory called RePower EU, which targets energy efficiency, renewable energy production and other areas related to environmental protection. Hence, this means that the EU has a long-term approach to helping build a sustainable ecosystem.

At the same time, our research suggests that European uniform environmental policies may be ineffective due to the significant differences in how regions experience the economic–environmental trade-off. For Western Europe, policies should focus on accelerating the transition to the declining segment of the EKC. At the same time, for Eastern Europe, more targeted measures may be needed to prevent the resurgence of emissions as economic development progresses.

Finally, it is necessary to acknowledge the specific constraints of our research regarding the limited size of the sample utilized. Although our findings offer significant insights, they may not apply to broader groups. Conducting future research with a broader sample could assist in overcoming this restriction and enhance the reliability of our findings. Additionally, further research may explore the reasons behind the diminishing long-term impact of renewable energy on environmental sustainability and investigate the potentially negative environmental effects of large-scale renewable projects and how to mitigate them. Also, studies may address why the combined effect of trade openness and renewables becomes insignificant for greenhouse emissions in the long term, identify factors contributing to this decreased effectiveness and explore potential policy interventions to maintain long-term benefits. Furthermore, investigations should examine the role of technological innovation in shaping the relationship between economic growth and environmental degradation. Besides, research on the optimal mix of policies (e.g., carbon pricing, regulations, incentives) that can effectively balance economic growth with environmental protection in different contexts may be a good option for advancing the current endeavour.