Green Transition in Europe: The Effectiveness of Environmental Taxes and Green Innovation in Reducing CO₂ Emissions

Abstract

This study examines the determinants of carbon dioxide (CO₂) emissions across 25 European Union countries over the period 2000–2021, with particular emphasis on the roles of environmental taxation and green innovation in shaping environmental sustainability. The analysis is grounded in ecological modernization theory, endogenous growth theory, and the Environmental Kuznets Curve hypothesis, which collectively explain the long-run and dynamic interactions between environmental policy, economic activity, structural transformation, and environmental outcomes. To ensure robust empirical inference, this study applies a comprehensive econometric framework that accounts for cross-sectional dependence, heterogeneity, non-stationarity, cointegration, and endogeneity. The empirical strategy begins with Pesaran cross-sectional dependence tests and slope heterogeneity diagnostics, followed by second-generation panel unit root tests (Pesaran CADF/CIPS) and Westerlund cointegration tests to establish the existence of long-run equilibrium relationships among the variables. Long-run coefficients are estimated using Fully Modified Ordinary Least Squares (FMOLS), Dynamic Ordinary Least Squares (DOLS), Canonical Cointegrating Regression (CCR), and Common Correlated Effects Mean Group (CCEMG) estimators. In addition, the Panel Autoregressive Distributed Lag (ARDL) model is employed to capture both short-run dynamics and long-run adjustment processes, while the System Generalized Method of Moments (System GMM) estimator addresses potential endogeneity, reverse causality, omitted variable bias, and dynamic persistence in CO₂ emissions. The empirical results indicate that environmental taxation has a positive and statistically significant association with CO₂ emissions, suggesting that current fiscal environmental policies in EU-25 countries may not yet be sufficiently effective in discouraging pollution-intensive activities. In contrast, green innovation is found to significantly reduce CO₂ emissions, underscoring the critical role of innovation-driven environmental investment and technological progress in improving environmental quality. Economic growth, exports, and urbanization are associated with higher emissions, while imports contribute to emission reductions, reflecting differences between domestic production-based effects and trade-related structural adjustments. The System GMM results further confirm the persistence of CO₂ emissions over time and validate the robustness of the long-run relationships identified by alternative estimators. Likewise, the CCEMG and Panel ARDL results support the stability and consistency of the findings under conditions of cross-sectional dependence and heterogeneous country dynamics. Taken together, the results highlight the importance of integrating environmental taxation with green innovation policies, innovation-driven investment, and sustainable trade policies to achieve long-term emission reductions in the European Union. This study contributes to the environmental economics literature by providing robust empirical evidence using second-generation panel econometric techniques that explicitly address cross-sectional dependence, heterogeneity, and endogeneity in the analysis of environmental sustainability.

1. Introduction

Climate change has emerged as one of the most critical global challenges, with significant implications for environmental sustainability, economic stability, and human welfare. Among greenhouse gases, carbon dioxide (CO₂) emissions remain the primary contributor to climate change, making emission reduction a central objective of international environmental policy (Nadiri et al., 2024; Ghazouani et al., 2021). In response, the European Union has adopted ambitious environmental strategies, including the European Green Deal and carbon-pricing initiatives, aimed at achieving climate neutrality by 2050 (Zahroh, 2025; Lobonţ et al., 2025). Nevertheless, persistent differences in economic structures, energy systems, and policy implementation continue to generate substantial variation in CO₂ emissions across EU member states (Mehta & Prajapati, 2024; Khan et al., 2023; Ghazouani et al., 2021). Consequently, identifying the policy and structural determinants of CO₂ emissions has become increasingly important within environmental economics research (Obobisa & Ahakwa, 2024; Nadiri et al., 2024; Khan et al., 2023; Q. Wang et al., 2023).

Recent environmental economics literature has increasingly shifted from traditional macroeconomic determinants toward policy-oriented and innovation-driven mechanisms capable of facilitating environmental sustainability. In this context, environmental taxation and green finance have become central instruments within the EU climate policy framework (Ghazouani et al., 2021; T. Wang et al., 2023). Unlike conventional economic variables, these policy instruments are specifically designed to influence environmental outcomes by altering production incentives, encouraging technological innovation, and supporting sustainable investments (Ghazouani et al., 2020; X. Wang et al., 2021). Environmental taxes aim to internalize environmental externalities by increasing the cost of pollution-intensive activities, thereby incentivizing firms and households to adopt cleaner technologies and improve energy efficiency (Ghazouani et al., 2021; Wolde-Rufael & Mulat-Weldemeskel, 2021; Zahroh, 2025; Soufiene et al., 2025). Within the EU, these taxes include carbon taxes, energy taxes, transport taxes, and pollution-related fiscal instruments that play an increasingly important role in environmental governance (Mehta & Prajapati, 2024; Lobonţ et al., 2025).

The theoretical foundation of environmental taxation is closely linked to Pigouvian taxation theory and the Porter Hypothesis, which suggest that appropriately designed environmental regulations can simultaneously improve environmental quality and stimulate technological innovation (Alola et al., 2022; Zahroh, 2025). Environmental taxes create market-based incentives that encourage firms to reduce emissions through technological upgrading, renewable energy adoption, and production restructuring (Nadiri et al., 2024; Khan et al., 2023; Koval et al., 2022; Lobonţ et al., 2025). Moreover, the “double dividend” hypothesis suggests that environmental taxation may generate both environmental improvements and economic efficiency gains if tax revenues are reinvested into sustainable infrastructure, innovation, and environmental protection (Soufiene et al., 2025). However, empirical evidence regarding the effectiveness of environmental taxes remains mixed. While numerous studies report that environmental taxation contributes to emission reduction (Nadiri et al., 2024; T. Wang et al., 2023; Ghazouani et al., 2021), other studies identify heterogeneous or nonlinear effects depending on income levels, industrial structures, energy dependency, and technological readiness (Khan et al., 2023; Gao & Fan, 2023; X. Wang et al., 2021; Soufiene et al., 2025).

Alongside environmental taxation, green innovation has become an increasingly important component of sustainable development strategies within the European Union. Green innovation broadly refers to public and private efforts directed toward environmentally sustainable activities, including renewable energy, sustainable infrastructure, environmental innovation, and low-carbon technologies (X. Wang et al., 2021; Ghazouani et al., 2021; Abderrahim & Benrekia, 2025; Kristkova et al., 2025). In the present study, green innovation is proxied by government environment-related research and development (R&D) expenditure as a percentage of GDP. However, it should be noted that this indicator does not represent financial intermediation or market-based green financing mechanisms such as green bonds or credit allocation. Instead, it captures a form of public sector innovation policy aimed at supporting environmental technology development and structural transformation. Therefore, its interpretation in this study should be understood as reflecting the role of government-led environmental innovation investment rather than conventional green innovation channels within financial markets (T. Wang et al., 2023). From a theoretical perspective, green innovation is strongly connected to endogenous growth theory, which emphasizes the role of innovation and technological progress in promoting long-run economic and environmental sustainability (Nadiri et al., 2024; X. Wang et al., 2021). Public investment in environmental R&D can reduce technological barriers, generate knowledge spillovers, and facilitate the transition toward cleaner production systems and renewable energy adoption (Ghazouani et al., 2020; Kristkova et al., 2025).

Within the EU framework, environmental taxation and green innovation are not isolated policy tools but rather complementary instruments capable of jointly supporting environmental sustainability objectives. Environmental taxes create economic incentives to reduce emissions, while green innovation provides the technological and policy capacity necessary to facilitate low-carbon transition processes (Kristkova et al., 2025). This interaction is particularly relevant in the European Union, where policy coordination and integrated environmental governance constitute central components of climate strategy (Lobonţ et al., 2025). Nevertheless, the extent to which environmental taxation and green innovation jointly influence CO₂ emissions remains insufficiently explored, particularly within highly integrated EU economies characterized by strong trade linkages, common environmental regulations, and coordinated climate policies.

This study makes several important contributions to the environmental economics literature.

First, it develops an integrated mechanism-based theoretical framework that jointly examines environmental taxation, green innovation, trade openness, urbanization, and economic growth within a unified setting for EU-25 countries over the period 2000–2021. Unlike much of the existing literature, which tends to analyze these determinants in isolation or within partial models, this study explicitly conceptualizes them as interconnected channels through which both policy instruments and structural economic factors shape environmental outcomes. In particular, it highlights the dual role of environmental taxation, which operates not only as a direct price-based mechanism for reducing emissions but also as an indirect driver of green innovation and structural economic adjustment. This integrated perspective allows for a more comprehensive understanding of how fiscal and innovation-oriented policies jointly contribute to long-run environmental performance.

Second, this study contributes methodologically by employing a rigorous and multi-layered econometric strategy designed to address key challenges in panel data analysis, including cross-sectional dependence, heterogeneity, non-stationarity, endogeneity, and dynamic persistence. By combining long-run estimators (FMOLS, DOLS, and CCR) with second-generation panel techniques (Pesaran CADF/CIPS tests, Westerlund cointegration tests, and CCEMG estimation), as well as dynamic approaches such as Panel ARDL and System GMM, this study ensures robustness from multiple econometric perspectives. This comprehensive approach strengthens the reliability of the estimated relationships and reduces concerns related to omitted variable bias, reverse causality, and unobserved common shocks, which are particularly relevant in cross-country environmental studies.

Third, this study provides new empirical insights into the heterogeneous and multi-channel nature of environmental sustainability determinants across EU countries. By incorporating trade openness and urbanization alongside fiscal and innovation-related variables, the analysis moves beyond aggregate macroeconomic explanations and captures structural differences in economic organization and development patterns. This allows for a more nuanced interpretation of how external trade structures, spatial economic concentration, and domestic policy instruments jointly influence CO₂ emissions. In doing so, this study highlights that environmental outcomes are not driven by single policies in isolation but by the interaction of multiple economic and institutional forces.

Fourth, this study contributes to the policy literature by offering evidence on the conditional effectiveness of environmental taxation and innovation-oriented policies within an integrated framework. The findings provide important implications for policy coordination in the European Union, suggesting that environmental taxation alone may not be sufficient unless complemented by strong innovation systems and structural transformation policies. Similarly, green innovation is shown to play a critical role in enhancing environmental performance, particularly when supported by appropriate fiscal incentives and broader economic reforms.

Finally, this study contributes by providing robust cross-country evidence from EU-25 economies over two decades, a relatively underexplored but highly relevant context characterized by strong institutional integration and policy harmonization. By combining advanced econometric techniques with an integrated theoretical framework, this study offers more reliable and policy-relevant insights into the complex drivers of CO₂ emissions and environmental sustainability.

2. Literature Review

The empirical literature on CO₂ emissions in Europe has grown substantially, with increasing attention given to the roles of economic growth, energy use, environmental taxation, green innovation, trade, and urbanization. Within this body of research, CO₂ emissions are commonly analyzed as a function of macroeconomic activity, policy instruments, and structural transformation processes. This section reviews the main findings from studies that examine the relationship between CO₂ emissions and key determinants, including GDP growth, environmental taxation, green innovation, exports, imports, and urbanization in EU countries.

2.1. Economic Growth and the Environmental Kuznets Curve

A robust finding in European studies is that economic growth tends to raise CO₂ emissions, at least up to certain income levels. For EU-27 over 2000–2017, Dynamic OLS estimates indicate that a 1% increase in GDP per capita raises CO₂ emissions by about 0.07%, confirming a positive long-run elasticity of emissions with respect to income (Onofrei et al., 2022). Similarly, spatial panel modeling for 26 EU countries (1990–2015) shows that GDP growth exerts positive total effects on domestic emissions in both the short and long run, even when spatial spillovers are accounted for (Ren et al., 2020).

Many EU studies explore the Environmental Kuznets Curve (EKC), an inverted U relationship between income and environmental degradation. Evidence of an EKC in European settings appears both for aggregate emissions and specific sectors. Panel analyses for new EU members and candidates (1992–2010) support an inverted U relationship between income and CO₂, consistent with the EKC hypothesis (Kasman & Duman, 2014). For broader European samples, economic growth initially increases transport sector CO₂ emissions, but EKC dynamics imply eventual decoupling at high income levels (A. Amin et al., 2020). Quantile panel regressions for 26 EU countries also confirm that economic growth has a positive effect on CO₂, with a squared term validating the EKC hypothesis (Khan et al., 2023).

Overall, the literature suggests that GDP growth is a core driver of emissions, but that at higher income levels, structural change, regulation, and technological progress can partly offset this impact (Neves et al., 2020; Khan et al., 2023; Onofrei et al., 2022; Kasman & Duman, 2014; Ren et al., 2020).

2.2. Environmental Taxes and Carbon Taxation

A central question for EU climate policy is whether environmental taxes effectively reduce CO₂ emissions. Several multi-country studies for Europe find that environmental taxes and carbon pricing are associated with lower emissions, though effects can be nonlinear.

Panel evidence for 20 European countries (1995–2012) shows that higher revenues from total environmental taxes, energy taxes, and transport taxes are significantly associated with lower CO₂ emissions; more stringent environmental policies also correlate with reduced emissions (Wolde-Rufael & Mulat-Weldemeskel, 2022). A dynamic panel threshold analysis for 15 EU members (1995–2013) reveals a nonlinear relationship: below threshold levels of environmental tax revenue (around 3% of GDP for total environmental taxes), effects on CO₂ are weak or even insignificantly positive, whereas, above the thresholds, environmental taxes significantly reduce emissions (Aydın & Esen, 2018). This suggests that the environmental effectiveness of taxation depends on the tax burden relative to GDP.

Recent work further documents asymmetries and dynamic effects. For the EU 27 over 1995–2022, the ARDL and NARDL models show that environmental taxes have a negative long-run impact on CO₂, with both symmetric and asymmetric relationships: increases in environmental taxes systematically reduce emissions, indicating that fiscal instruments successfully disincentivize polluting activities (Mehta & Prajapati, 2024).

Carbon-specific taxation also matters. A study of residential CO₂ emissions in 19 European countries (2000–2017) finds that a €20-per-tonne CO₂ tax reduces residential emissions by about 1% on average, with heterogeneous tax revenue burdens across countries (Charlier et al., 2023). Another panel for EU members indicates that carbon taxes, coupled with eco innovation and renewable energy deployment, slowed environmental degradation over 1994–2019 (Nadiri et al., 2024).

More recent cross-European work explicitly combines environmental tax with green innovation and finance. For 25 European countries (1990–2019), environmental tax is found to significantly reduce CO₂ emissions, in line with the EU’s net zero ambitions, while GDP per capita and non-renewable energy use increase emissions (Obobisa & Ahakwa, 2024). Quantile on quantile regressions suggest that the mitigating effect of environmental tax persists across the emissions distribution (Obobisa & Ahakwa, 2024).

In sum, there is a growing consensus that environmentally related taxes, when set at sufficiently high levels and embedded in broader policy packages, contribute to CO₂ abatement in EU countries (Aydın & Esen, 2018; Nadiri et al., 2024; Charlier et al., 2023).

2.3. Green Finance, Green Innovation, and Environmental R&D

Green innovation and environment-related R&D are widely recognized in the literature as important mechanisms for supporting low-carbon investment, technological progress, and environmental sustainability. While empirical studies specifically focusing on EU countries remain relatively limited, existing cross-country evidence suggests that public and private investments in environmental innovation can play a significant role in reducing emissions and facilitating the transition toward cleaner production systems.

For 21 European countries, cross-sectional ARDL estimates indicate that green technology innovation and renewable energy usage reduce CO₂ emissions, whereas financial development and GDP increase them (Zhou et al., 2024). Human capital further enhances the emissions-reducing effects of green innovation (Zhou et al., 2024).

A broader sample of 25 “environmentally friendly” countries (EPI leaders) for 2000–2021 finds that green finance, green innovation, green growth, environmental taxes, urbanization, and trade openness all exert negative effects on CO₂ emissions across the distribution of emissions, using movement quantile regression and standard quantile regression (Wei et al., 2025). These results imply that green financial flows and innovation, when combined with strong policy frameworks, reinforce environmental quality at low, median, and high emission levels.

Outside Europe, but conceptually relevant, panel quantile regressions for G20 economies (2008–2018) show that green finance, renewable energy investment, and technological innovation lower CO₂ emissions, whereas economic growth, energy consumption, trade, and FDI increase them (Zhang et al., 2022). Conditional process analysis for Chinese provinces reveals that green finance can directly suppress CO₂ emissions and indirectly do so by fostering technological innovation; green fiscal policy (including green taxes) strengthens these indirect effects (Hu et al., 2023). These findings support the theoretical premise that environment-related public R&D and dedicated green finance channels are important determinants of emissions through innovation-driven efficiency gains and diffusion of clean technologies (Zhang et al., 2022; Zhou et al., 2024; Hu et al., 2023).

For a study that proxies green innovation by environment-related government R&D spending, this literature suggests expecting a negative long-run relationship between environmental R&D (as % of GDP) and CO₂ emissions, especially when R&D is embedded in broader green growth strategies.

2.4. International Trade and CO₂ Emissions

The role of trade openness in emissions is multifaceted, involving scale, composition, and technique effects. For new EU members and candidate countries (1992–2010), panel causality tests show that trade openness Granger causes CO₂ emissions in the short run, supporting a scale effect in which higher trade volumes are associated with increased emissions (Kasman & Duman, 2014).

However, more recent work for the EU finds differentiated effects of exports and imports. Wavelet analysis on the trade-CO₂ nexus in the EU (1960–2014) shows that CO₂ emissions drive trade in the short and long run, under the influence of economic shocks, energy use and growth; in the medium term, imports drive emissions via a scale effect, while in the long run an “inverted pollution haven” effect emerges as exports are increasingly associated with cleaner capacities (Mutascu & Sokic, 2020).

A separate trade gravity study documents that EU exports and imports tend to grow toward partner countries whose CO₂ emissions are rising, reflecting both the export of environmental and high-tech goods and the offshoring of emission-intensive production (Steinhauser et al., 2024). Evidence from firm-level and sectoral studies (cited therein) supports the view that exporting firms tend to be less emission-intensive, while high-emission activities may be displaced abroad (Steinhauser et al., 2024).

Other EU-focused work finds that increasing exports is associated with falling domestic CO₂ emissions, whereas higher imports correlate with emission increases, consistent with the idea that export-oriented sectors become cleaner while import patterns embody carbon-intensive production overseas (Steinhauser et al., 2024). In a broader green transition sample, trade openness appears to have a negative association with CO₂ when combined with green finance, innovation, and environmental taxes, suggesting that openness under stringent environmental regimes may reinforce decarbonization (Wei et al., 2025).

For a model distinguishing exports and imports (both as % of GDP), the literature therefore suggests potentially opposite signs: exports may coincide with improved emission intensity, whereas imports can be associated with higher embodied emissions, depending on partner composition and policy context (Steinhauser et al., 2024; Kasman & Duman, 2014; Mutascu & Sokic, 2020).

2.5. Urbanization and CO₂ Emissions

Urbanization is another key determinant of emissions, but its sign is empirically ambiguous. For a panel of EU countries (2000–2018), cointegration analysis finds a long-run negative impact of urbanization on CO₂ emissions per capita: higher urbanization rates are associated with lower emissions, contrary to many earlier findings (Jóźwik et al., 2022). The authors attribute this to more energy-efficient infrastructure, technological progress, and environmental policies in EU cities, suggesting that advanced urbanization may support decarbonization (Jóźwik et al., 2022).

In contrast, earlier work on Southern and Central Eastern European countries typically finds that urbanization increases emissions. For example, studies summarized in Jóżwik et al. indicate that in Portugal, Ireland, Italy, Greece, and Spain, a 1% rise in urbanization increased CO₂ emissions by 0.44-6.36% depending on the model (Jóźwik et al., 2022). Analysis of new EU members and candidate states (1992–2010) also finds short-run unidirectional causality from urbanization to CO₂ emissions, implying that urban expansion initially raises environmental pressures (Kasman & Duman, 2014).

Sectoral evidence for European transport shows that urbanization exerts a (statistically insignificant) positive impact on transport sector CO₂, while renewable energy in transport significantly reduces emissions (A. Amin et al., 2020). The mixed results imply that the impact of urbanization may be nonlinear and stage-dependent, with early urban growth increasing energy demand and later stages enabling efficiency gains and low-carbon infrastructure (Jóźwik et al., 2022; A. Amin et al., 2020; Kasman & Duman, 2014).

Studies incorporating urbanization into broader green finance/green growth frameworks find that urbanization can have a negative association with emissions when combined with strong environmental policies, green technologies, and sustainable urban planning (Wei et al., 2025). This suggests that the sign of urbanization in empirical models for EU 25 is likely to depend on the time period, level of development, and interaction with policy instruments.

2.6. Energy Structure, Environmental Regulation, and Spatial Spillovers

Renewable energy use in European countries consistently shows negative effects on CO₂ emissions, while non-renewable energy use and total energy consumption increase emissions (Obobisa & Ahakwa, 2024; Khan et al., 2023; A. Amin et al., 2020; Ren et al., 2020; Nazarko et al., 2022). For example, analysis of transport sector emissions in Europe finds that increasing renewable energy consumption reduces transport emissions by about 12%, validating a strong mitigation effect (A. Amin et al., 2020).

Environmental regulation, more broadly, beyond taxation, also matters. For 17 EU countries (1995–2017), an ARDL model with Driscoll-Kraay standard errors shows that market-based environmental regulations and policies supporting renewable energy reduce CO₂ emissions in the long run, and that foreign direct investment has a “pollution halo” effect, implying that the EU increasingly attracts cleaner, innovative investment (Neves et al., 2020).

Spatial econometric work on 26 EU countries highlights positive spatial spillovers of CO₂ emissions, meaning that emissions in one country are correlated with those of its neighbors (Ren et al., 2020). Economic growth has positive domestic effects on emissions, but negative spatial spillovers, implying that growth in neighboring countries may reduce local emissions through technology diffusion or regulatory competition (Ren et al., 2020). These findings underscore the importance of considering cross-border interactions, regional policy coordination, and diffusion of green technologies in EU-wide analyses.

2.7. Research Hypotheses

Drawing on theoretical insights and empirical evidence from the environmental economics and sustainability literature, this study formulates a set of testable hypotheses concerning the determinants of CO₂ emissions in EU-25 countries. Particular emphasis is placed on policy-related variables, namely, environmental taxation and green innovation, while also accounting for key structural and macroeconomic factors.

Hypothesis 1. 

Economic growth is expected to have a positive effect on CO₂ emissions, as higher GDP growth is typically associated with increased industrial output, energy consumption, and resource utilization. However, this relationship may exhibit nonlinearity consistent with the Environmental Kuznets Curve (EKC) hypothesis. Specifically, emissions are expected to rise at lower levels of economic development but decline beyond a certain income threshold due to structural transformation, technological progress, and stricter environmental regulations (Khan et al., 2023; Onofrei et al., 2022; Kasman & Duman, 2014; Ren et al., 2020).

Hypothesis 2. 

Environmental taxes are hypothesized to reduce CO₂ emissions by internalizing the negative externalities associated with pollution. By increasing the cost of carbon-intensive activities, such taxes create incentives for firms and households to adopt cleaner production processes and energy sources. The effectiveness of environmental taxation is expected to strengthen as tax levels increase, suggesting a potentially nonlinear relationship and more pronounced long-run effects (Obobisa & Ahakwa, 2024; Mehta & Prajapati, 2024; Wolde-Rufael & Mulat-Weldemeskel, 2022; Aydın & Esen, 2018; Nadiri et al., 2024; Charlier et al., 2023).

Hypothesis 3. 

Green innovation, proxied by government expenditure on environment-related research and development as a percentage of GDP, is expected to have a negative impact on CO₂ emissions. Increased public investment in environmental R&D fosters technological innovation, supports the development of cleaner production systems, and reduces the carbon intensity of economic activity. These effects are likely to be strengthened when green innovation is supported by complementary policy frameworks, reinforcing its role in long-term emission reduction (Wei et al., 2025; Zhang et al., 2022; Zhou et al., 2024; Hu et al., 2023).

Hypothesis 4. 

Exports are expected to be associated with a reduction in CO₂ emission intensity. In the context of EU economies, export-oriented sectors are often characterized by higher productivity, technological advancement, and stricter environmental standards. As a result, increased exports may reflect a shift toward cleaner production processes and improved energy efficiency, thereby contributing to lower emissions (Steinhauser et al., 2024; Kasman & Duman, 2014).

Hypothesis 5. 

Imports are expected to have a positive effect on CO₂ emissions. Imported goods may embody carbon-intensive production processes from external economies and contribute to higher consumption-based emissions within importing countries. This effect reflects the potential displacement of pollution through international trade, whereby emissions are embedded in imported products (Mutascu & Sokic, 2020; Steinhauser et al., 2024).

Hypothesis 6. 

The effect of urbanization on CO₂ emissions is expected to be ambiguous. On the one hand, urbanization can increase emissions through higher energy demand, transportation needs, and infrastructure expansion. On the other hand, it can promote efficiency gains through economies of scale, improved public transportation systems, and more effective environmental governance. In the EU context, where urban systems are relatively advanced, the net effect may be neutral or even negative (Jóźwik et al., 2022; A. Amin et al., 2020; Kasman & Duman, 2014).

These insights suggest modeling potential nonlinear and asymmetric effects (e.g., thresholds for environmental taxes, EKC for GDP, quantile behavior), and considering the interaction between green innovation, environmental taxation, and other structural variables when explaining CO₂ emissions in the EU-25 countries.

3. The Theoretical Framework

The theoretical framework of this study is grounded in Ecological Modernization Theory, the Porter Hypothesis, and endogenous growth theory, which collectively provide a comprehensive explanation of how environmental policy instruments and sustainable investment channels shape environmental quality over time. These theories jointly suggest that economic development and environmental protection are not necessarily conflicting objectives, but can become mutually reinforcing through appropriate policy design, innovation, and structural transformation.

Within this framework, CO₂ emissions are determined by a combination of economic activity, environmental taxation, green innovation, trade dynamics, and urbanization. Environmental taxation plays a central role by internalizing environmental externalities and increasing the cost of pollution-intensive production and consumption. This price-based mechanism encourages firms and households to shift toward cleaner production processes, improve energy efficiency, and reduce reliance on carbon-intensive inputs.

Green innovation complements this mechanism by addressing the technological constraints of environmental transition. Public investment in environmental R&D promotes knowledge creation, supports technological diffusion, and accelerates the development of low-carbon technologies. In line with endogenous growth theory, such innovation-driven processes generate long-run improvements in productivity while simultaneously reducing the carbon intensity of economic activity.

At the same time, economic growth influences environmental outcomes through scale and composition effects. While higher income levels initially tend to increase emissions due to expanded production and energy demand, they may also create conditions for cleaner technologies and stronger environmental regulations in later stages of development. International trade further shapes emissions through specialization patterns, production relocation, and differences in environmental standards across countries. Finally, urbanization affects environmental quality by concentrating economic activity, increasing energy demand in transport and housing, and altering infrastructure and consumption patterns, while also potentially generating efficiency gains through scale economies and improved service provision. Finally, the framework emphasizes that environmental degradation is the result of interacting economic, policy, and structural forces rather than a single determinant, highlighting the importance of integrated policy approaches in achieving long-term environmental sustainability.

The environmental degradation function can therefore be expressed as:

$${CO}_{2it}=f(E{T}_{it},{GI}_{it},GD{P}_{it},IMPOR{T}_{it},EXPOR{T}_{it},URBA{N}_{it})$$

(1)

where ${CO}_{2it}$ denotes carbon dioxide emissions for country $i$ at time $t$, $E{T}_{it}$ represents environmental taxation, ${GI}_{it}$ denotes green innovation, $GD{P}_{it}$ captures economic growth, $IMPOR{T}_{it}$ and $EXPOR{T}_{it}$ represent trade activities, and $URBA{N}_{it}$ measures urbanization.

4. Data and Methodology

4.1. Data

This study employs a balanced panel dataset covering 25 European Union (EU-25) countries over the period 2000–2021. The choice of this sample reflects both data availability and the policy relevance of these economies, particularly in the context of environmental taxation and green innovation. The panel structure enables the analysis to capture cross-country heterogeneity as well as temporal dynamics in the determinants of CO₂ emissions.

Variables and Measurement

Table 1. Variables and data sources.

Variable	Description	Measurement	Year	Source
CO2	Carbon dioxide emissions	CO2 emissions (total) (% change from 1990)	2000–2021	World Bank (WDI)
gdp	Economic growth	Annual % growth rate of GDP	2000–2021	World Bank (WDI)
et	Environmental tax revenue	% of GDP	2000–2021	International Monetary Fund (IMF)
gi	Environment-related government R&D expenditure	% of GDP	2000–2021	Organisation for Economic Co-operation and Development (OECD)
export	Exports of goods and services	% of GDP	2000–2021	World Bank (WDI)
import	Imports of goods and services	% of GDP	2000–2021	World Bank (WDI)
urban	Urban population	% of total population	2000–2021	World Bank (WDI)

Note: CO₂ = carbon dioxide emissions; gdp = economic growth rate; et = environmental tax revenue; gi = green innovation (proxied by government environment-related R&D expenditure); export = exports of goods and services; import = imports of goods and services; urban = urbanization rate.

shows the variables utilezed in this study. The dependent variable in this study is CO₂ emissions, which serve as a proxy for environmental degradation. CO₂ emissions are measured as total CO₂ emissions (% change from 1990) and reflect changes in carbon intensity associated with economic activity over time. The primary explanatory variables of interest are environmental taxation and green innovation. Environmental tax is measured as total environmental tax revenue expressed as a percentage of GDP. This variable captures the extent to which governments rely on fiscal instruments to internalize environmental externalities and discourage environmentally harmful activities. A higher value indicates a stronger policy orientation toward market-based environmental regulation. Green innovation is proxied by government expenditure on environment-related R&D as a percentage of GDP. This measure reflects the role of public investment in supporting environmental innovation, technological advancement, and the transition toward a low-carbon economy. By focusing on environment-related R&D, this study captures the long-term structural dimension of green innovation. In addition to the core variables, the analysis incorporates a set of control variables to account for broader macroeconomic and structural influences on CO₂ emissions. Economic growth is measured as the annual percentage growth rate of GDP. Exports and imports are expressed as percentages of GDP, allowing for a clear distinction between production-oriented and consumption-driven environmental effects. Urbanization is measured as the percentage of the population residing in urban areas, capturing demographic and structural changes that may influence energy consumption and environmental outcomes.

The dataset utilized in this study is compiled from multiple internationally recognized databases to ensure a high degree of reliability, consistency, and cross-country comparability over time. Specifically, data on CO₂ emissions, GDP growth, exports, imports, and urbanization are obtained from the World Bank’s World Development Indicators (WDI), which provides standardized and widely used macroeconomic and environmental indicators. Environmental tax data are sourced from the International Monetary Fund (IMF), offering comprehensive and harmonized fiscal statistics, including environmental tax revenues expressed as a percentage of GDP. In addition, data on green innovation, proxied by government expenditure on environment-related research and development, are drawn from the Organisation for Economic Co-operation and Development (OECD), which supplies detailed and consistent information on R&D activities across countries.

Table 2. Descriptive statistics.

Variable	Mean	Std. Dev.	Min	Max
CO2	−10.7015	29.35874	−69.69813	63.59207
et	2.655836	0.6948871	1.13	5.1
gi	2.702863	1.84178	0.05	10.46
gdp	2.370118	3.865162	−16.04002	24.62399
import	5.196319	8.60034	−30.8932	41.29586
export	5.245952	7.777822	−30.34822	39.42661
urban	71.55108	12.02323	49.76195	94.62948

presents the descriptive statistics of the variables employed in the analysis for the EU-25 countries over the period 2000–2021. The results indicate substantial variation across countries and over time, reflecting differences in environmental policies, economic structures, trade integration, energy transition processes, and demographic characteristics within the European Union. The dependent variable, CO₂ emissions measured as the percentage change from 1990, records an average value of −10.70, suggesting that EU-25 countries, on average, achieved reductions in emissions relative to the 1990 baseline during the study period. However, the relatively large standard deviation of 29.36 indicates considerable heterogeneity in environmental performance across countries. The minimum value of −69.70 and the maximum value of 63.59 further demonstrate that, while some countries experienced substantial reductions in emissions, others recorded significant increases over time.

Environmental taxation exhibits an average value of 2.66% of GDP, with values ranging from 1.13% to 5.10% of GDP, indicating differences in the intensity and implementation of environmental fiscal policies across EU member states. The relatively moderate standard deviation suggests that environmental tax systems are comparatively stable within the region. Green innovation records an average value of 2.70 with a standard deviation of 1.84, indicating considerable variation in public environmental R&D efforts across EU countries. The relatively wide dispersion between minimum and maximum values suggests significant heterogeneity in government commitment to environmental innovation, with some countries investing substantially in green R&D and others allocating comparatively limited resources. Economic growth, measured by the annual GDP growth rate, has an average value of 2.37%, indicating moderate economic expansion across the EU-25 during the study period. The relatively high standard deviation of 3.87 highlights fluctuations in economic performance across countries and time periods, particularly during periods of economic instability, such as the global financial crisis. The negative minimum value additionally reflects episodes of economic contraction experienced by several countries. Imports and exports exhibit average values of approximately 5.20% and 5.25%, respectively, suggesting the high degree of trade openness characterizing EU economies. The relatively large standard deviations and wide ranges of both trade variables indicate substantial volatility in international trade activities across countries and over time. Urbanization records the highest average among the variables, with approximately 71.55% of the population residing in urban areas, confirming the highly urbanized nature of EU economies. The comparatively lower standard deviation indicates that urbanization levels are relatively stable across the region, although some variation remains among member states. Overall, the descriptive statistics reveal substantial heterogeneity in environmental, economic, trade, and demographic conditions across EU-25 countries, thereby justifying the application of advanced panel econometric techniques capable of accounting for heterogeneity, dynamic relationships, and cross-sectional dependence within the dataset.

Table 3. Correlation matrix and VIF test.

Variables	CO2	et	re	gdp	Import	Export	Urban
CO2	1.000
et	0.512	1.000
gi	−0.438	−0.221	1.000
gdp	0.483	0.371	−0.297	1.000
import	−0.362	−0.194	0.211	0.402	1.000
export	0.416	0.288	−0.186	0.533	0.624	1.000
urban	0.274	0.156	−0.139	0.417	0.288	0.301	1.000
Variables	VIF	1/VIF
export	3.28	0.305
import	3.11	0.321
gdp	2.84	0.352
et	2.21	0.452
gi	1.93	0.518
urban	1.67	0.599
Mean VIF	2.51	-

reports the correlation matrix and Variance Inflation Factor (VIF) results for the variables used in the analysis of CO₂ emissions in EU-25 countries over the period 2000–2021. The correlation results show that CO₂ emissions are positively associated with environmental taxation, economic growth, exports, and urbanization, while imports and green innovation (environment-related R&D expenditure) are negatively correlated with emissions. In particular, the negative correlation between green innovation and CO₂ emissions highlights the potential role of public environmental R&D in supporting cleaner production and environmental improvement across EU countries. In contrast, the positive association between economic growth and emissions suggests that economic expansion continues to exert upward pressure on environmental degradation despite policy efforts.

The VIF results indicate no serious multicollinearity concerns among the explanatory variables. All VIF values are well below the conventional threshold of 10, with a mean VIF of 2.51, confirming that the regressors can be jointly included in the model without causing significant estimation bias or instability. Overall, these diagnostic results support the suitability and reliability of the variables for subsequent panel econometric analysis.

Table 4. Pesaran Cross-Sectional Dependence (CD) test.

Variables	CD Statistic	p-Value
CO2	18.742 ***	0.000
et	11.584 ***	0.000
gi	9.213 ***	0.000
gdp	15.826 ***	0.000
import	12.771 ***	0.000
export	13.445 ***	0.000
urban	7.936 ***	0.000

Notes: *** indicates significance at the 1% level.

reports the results of the Pesaran Cross-Sectional Dependence test for the variables included in the analysis. The findings reveal that the null hypothesis of cross-sectional independence is strongly rejected for all variables at the 1% significance level. The statistically significant CD statistics indicate the presence of substantial cross-sectional dependence among EU-25 countries over the study period. This outcome is expected given the high degree of economic integration within the European Union through common environmental regulations, trade linkages, energy markets, monetary coordination, and regional policy frameworks.

The presence of cross-sectional dependence implies that shocks affecting one EU country may transmit to other member states through economic and institutional channels. Consequently, the use of conventional first-generation panel econometric techniques that assume cross-sectional independence may lead to biased and inconsistent results. These findings therefore justify the application of second-generation panel unit root and cointegration tests, as well as advanced estimators such as CCEMG and System GMM, which explicitly account for cross-sectional dependence and heterogeneity across countries.

4.2. Methodology

4.2.1. Model Specification

To examine the determinants of CO₂ emissions in EU-25 countries, this study specifies a panel data model incorporating environmental taxation, green innovation, economic growth, trade variables, and urbanization. The empirical framework is grounded in ecological modernization theory, endogenous growth theory, and the Environmental Kuznets Curve hypothesis, which collectively explain how environmental policy instruments, technological transformation, economic expansion, and structural changes influence environmental quality.

The functional relationship can be expressed as follows:

$${CO}_{2it}=f(E{T}_{it},{GI}_{it},GD{P}_{it},IMPOR{T}_{it},EXPOR{T}_{it},URBA{N}_{it})$$

(2)

To obtain an estimable econometric specification, the model is expressed in its linear panel form:

$$\begin{gathered} {CO}_{2it}=\alpha +{\beta }_{1}E{T}_{it}+{\beta }_2{GI}_{it}+{\beta }_{3}GD{P}_{it}+{\beta }_{4}IMPOR{T}_{it}+{\beta }_{5}EXPOR{T}_{it} \\ +{\beta }_{6}URBA{N}_{it}+{\epsilon }_{it} \end{gathered}$$

(3)

where $i=\mathrm{1,2},\dots ,25$ denotes cross-sectional units and $t=2000,\dots ,2021$ represents the time dimension. ${\epsilon }_{it}$ is the stochastic error term. The dependent variable, ${CO}_2$, represents carbon dioxide emissions measured in % change from 1990 obtained from the World Development Indicators database. Environmental taxation ($ET$) represents environmentally related tax revenues as a percentage of GDP, while green innovation ($GI$) captures green innovation. Economic growth ($GDP$) is measured using GDP per capita, whereas imports and exports are expressed as percentages of GDP to capture the environmental implications of international trade activities. Urbanization ($URBAN$) measures the proportion of the urban population.

All variables are used in their original form without logarithmic transformation. This specification is adopted to maintain consistency with the percentage-based and growth-rate nature of the dataset, particularly for variables expressed as percentage changes or shares of GDP. The inclusion of both environmental taxation and green innovation (proxied by environment-related R&D expenditure) allows the model to capture the complementary roles of environmental policy instruments and public innovation efforts in shaping environmental outcomes and supporting the transition toward lower emissions.

4.2.2. Preliminary Econometric Diagnostics

Given the strong economic, institutional, and environmental integration among EU-25 countries through common energy markets, environmental regulations, trade linkages, and EU-wide climate policies, cross-sectional dependence is likely to exist within the panel dataset. Ignoring cross-sectional dependence may lead to biased and inconsistent estimation results. Therefore, this study first conducts the Pesaran Cross-Sectional Dependence test to examine the presence of cross-sectional correlation among panel units.

Furthermore, slope heterogeneity tests are performed to determine whether the slope coefficients differ across countries. Since EU economies exhibit different economic structures, environmental policies, and energy consumption patterns, assuming homogeneous slope coefficients may not be appropriate.

Following the diagnostic tests, second-generation panel unit root tests are employed to determine the stationarity properties of the variables while accounting for cross-sectional dependence. Specifically, the Pesaran CADF/CIPS unit root tests are applied because they provide more reliable results than first-generation tests in the presence of cross-sectional correlation. Establishing the order of integration is essential to avoid spurious regression problems and to determine the appropriate econometric framework for estimation.

After confirming the integration properties of the variables, the Westerlund panel cointegration test is employed to examine the existence of long-run equilibrium relationships among the variables. The Westerlund approach is preferred because it accommodates cross-sectional dependence and heterogeneity across countries. The confirmation of cointegration justifies the use of long-run panel estimators and indicates that the variables move together over time despite short-run fluctuations.

4.2.3. Long-Run Estimation Techniques

To estimate the long-run relationships among the variables, this study employs Fully Modified Ordinary Least Squares, Dynamic Ordinary Least Squares, and Canonical Cointegrating Regression estimators. The use of multiple estimators enhances the robustness and reliability of the empirical findings by allowing cross-validation of long-run coefficient estimates.

The FMOLS estimator corrects for serial correlation and endogeneity arising from cointegrated regressors through non-parametric adjustments. The FMOLS estimator can be expressed as:

$${\widehat{\beta }}_{FMOLS}={\left({\displaystyle {\sum }_{t=1}^T}(X_t-\overline{X}\left)\right(X_t-\overline{X}{)}^{\prime }\right)}^{-1}\left({\displaystyle {\sum }_{t=1}^T}(X_t-\overline{X}\left)\right(Y_t-\overline{Y}{)}^{*}\right)$$

(4)

where ${\left(Y_t-\overline{Y}\right)}^{*}$ represents the transformed dependent variable adjusted for serial correlation and endogeneity using long-run covariance matrices.

The DOLS estimator addresses potential endogeneity by augmenting the regression with leads and lags of the first differences of the explanatory variables. The DOLS specification is expressed as:

$${CO}_{2it}={\alpha }_i+\beta X_{it}+{\displaystyle {\sum }_{k=-q}^q}{\gamma }_k\Delta X_{it-k}+{\epsilon }_{it}$$

(5)

where $X_{it}$ represents the vector of explanatory variables and $\Delta X_{it-k}$ denotes the leads and lags of first differences included to eliminate feedback effects and improve estimation efficiency.

Similarly, the CCR estimator transforms the variables to eliminate long-run correlation between the regressors and the error term. The CCR transformation is given by:

$$Y_t^{*}=Y_t-{\widehat{\mathsf{\Omega }}}_{21}{\widehat{\mathsf{\Omega }}}_{22}^{-1}{X}_t$$

(6)

$$X_t^{*}=X_t-{\widehat{\mathsf{\Omega }}}_{22}^{-1}{\widehat{\mathsf{\Omega }}}_{21}^{\prime }$$

(7)

where ${\widehat{\mathsf{\Omega }}}_{21}$ and ${\widehat{\mathsf{\Omega }}}_{22}$ are elements of the long-run covariance matrix. The transformed variables are then estimated using ordinary least squares to obtain efficient long-run coefficients.

The combined use of FMOLS, DOLS, and CCR improves the reliability of the empirical analysis by addressing serial correlation and partial endogeneity problems associated with cointegrated panel data models.

4.2.4. Dynamic Panel Estimation: System GMM

Although FMOLS, DOLS, and CCR partially address endogeneity associated with cointegrated regressors, they may not fully resolve deeper structural endogeneity problems arising from reverse causality, simultaneity bias, omitted variables, and dynamic persistence. Environmental taxation may react to rising emission levels, renewable energy investment may increase following environmental degradation, and economic growth and trade activities may simultaneously influence and be influenced by CO₂ emissions.

To address these concerns, this study additionally employs the System Generalized Method of Moments estimator developed by Arellano and Bover and further extended by Blundell and Bond. The System GMM framework is particularly appropriate for dynamic macro-panel analysis because it effectively controls for several important econometric problems commonly observed in environmental panel data studies. Specifically, the estimator addresses potential endogeneity arising from simultaneity and feedback effects between CO₂ emissions and the explanatory variables, while also mitigating reverse causality issues whereby environmental taxation, renewable energy investment, and trade patterns may themselves respond to environmental degradation. In addition, the System GMM estimator helps reduce omitted variable bias by controlling for unobserved country-specific effects and incorporates the dynamic persistence of CO₂ emissions through the inclusion of the lagged dependent variable. The estimator therefore provides more reliable and consistent parameter estimates in the presence of endogenous regressors, serial correlation, and dynamic panel bias, making it particularly suitable for the analysis of environmental sustainability in EU-25 countries.

The dynamic specification is expressed as follows:

$$\begin{gathered} {CO}_{2it}=\alpha +\rho {CO}_{2it-1}+{\beta }_{1}E{T}_{it}+{\beta }_2{GI}_{it}+{\beta }_{3}GD{P}_{it}+{\beta }_{4}IMPOR{T}_{it} \\ +{\beta }_{5}EXPOR{T}_{it}+{\beta }_{6}URBA{N}_{it}+{\epsilon }_{it} \end{gathered}$$

(8)

where $\mathrm{l}\mathrm{n} {CO}_{2it-1}$ represents the lagged dependent variable capturing the persistence of environmental degradation over time.

Lagged levels and lagged first differences of the endogenous variables are employed as internal instruments because they are strongly correlated with current values while remaining uncorrelated with the contemporaneous error term. The validity of the instruments is evaluated using the Hansen J-test, while the Arellano-Bond AR(1) and AR(2) tests are used to examine serial correlation in the residuals.

4.2.5. Common Correlated Effects Mean Group Estimator

To further account for cross-sectional dependence and slope heterogeneity among EU-25 countries, this study additionally employs the Common Correlated Effects Mean Group estimator proposed by Pesaran. The CCEMG estimator incorporates cross-sectional averages of the dependent and independent variables to control for unobserved common factors shared across countries, including common environmental regulations, energy market shocks, EU climate policies, and regional economic integration.

The CCEMG specification can be expressed as:

$${CO}_{2it}={\alpha }_i+{\beta }_i{X}_{it}+{\gamma }_i{\overline{Y}}_t+{\delta }_i{\overline{X}}_t+{\epsilon }_{it}$$

(9)

where ${\overline{Y}}_t$ and ${\overline{X}}_t$ represent the cross-sectional averages of the dependent and explanatory variables, respectively.

The inclusion of CCEMG estimation strengthens the robustness of the empirical findings by addressing the limitations associated with conventional panel estimators under cross-sectional dependence.

4.2.6. Panel ARDL Approach and Robustness Analysis

To further validate the empirical findings and simultaneously capture short-run dynamics and long-run equilibrium relationships, this study employs the Panel Autoregressive Distributed Lag model as an additional robustness estimation technique. The Panel ARDL framework is advantageous because it can accommodate variables integrated of different orders, provided that none are integrated of order two.

The Panel ARDL error correction specification is expressed as follows:

$$\Delta {CO}_{2it}=\varphi ({CO}_{2it-1}-{\theta }_{1}E{T}_{it-1}-{\theta }_2{GI}_{it-1}-\cdots )+\sum {\gamma }_j\Delta X_{it-j}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(10)

where $\varphi$ represents the speed of adjustment toward long-run equilibrium, $\Delta$ denotes first differences, and the error correction term captures deviations from the long-run equilibrium path.

The Panel ARDL approach provides several advantages. First, it allows simultaneous estimation of short-run and long-run effects. Second, it accommodates heterogeneity across countries through country-specific dynamics. Third, it serves as an additional robustness check for the long-run estimators and dynamic panel estimations, thereby strengthening confidence in the stability and consistency of the empirical findings.

4.2.7. Unit-Root Tests

To examine the stationarity properties of the variables while accounting for cross-sectional dependence among EU-25 countries, this study employs the Pesaran Cross-sectionally Augmented Dickey-Fuller and Cross-sectionally Im-Pesaran-Shin panel unit root tests over the period 2000–2021. The use of second-generation panel unit root tests is particularly appropriate because EU economies are highly interconnected through common environmental regulations, energy markets, trade integration, financial linkages, and macroeconomic coordination. Conventional first-generation panel unit root tests assume cross-sectional independence and may therefore generate biased and inconsistent results in the presence of common shocks and regional interdependence. The Pesaran CADF/CIPS approach overcomes this limitation by augmenting standard unit root regressions with cross-sectional averages, thereby producing more reliable stationarity results under cross-sectional dependence.

Table 5. Pesaran CADF/CIPS unit root test.

Variables	CADF Level	CADF First Difference	CIPS Level	CIPS First Difference	Order of Integration
CO2	−1.764	−5.182 ***	−1.842	−4.926 ***	I(1)
et	−1.913	−5.437 ***	−2.014	−5.113 ***	I(1)
gi	−2.104	−4.886 ***	−1.931	−4.882 ***	I(1)
gdp	−1.887	−5.621 ***	−2.107	−5.341 ***	I(1)
import	−1.736	−4.744 ***	−1.764	−4.567 ***	I(1)
export	−1.845	−4.918 ***	−1.886	−4.731 ***	I(1)
urban	−1.992	−5.247 ***	−2.038	−5.014 ***	I(1)

Notes: *** indicates significance at the 1% level. CADF refers to the Cross-sectionally Augmented Dickey-Fuller test, while CIPS denotes the Cross-sectionally Im-Pesaran-Shin panel unit root test developed by Pesaran (2007). Both tests account for cross-sectional dependence among panel units. The results indicate that all variables become stationary after first differencing and are therefore integrated of order one, I(1).

presents the results of the Pesaran (2007) CADF and CIPS panel unit root tests for all variables. The null hypothesis of these tests assumes the presence of a unit root, while the alternative hypothesis indicates stationarity. The decision regarding stationarity is based on whether the test statistics exceed the relevant critical values in absolute (more negative) terms, rather than on the sign or magnitude of the coefficients themselves. Accordingly, a variable is considered stationary only when the CADF/CIPS statistics are sufficiently more negative than the corresponding critical values. At levels, although the CADF and CIPS statistics appear negative, they do not reach the critical thresholds required to reject the null hypothesis of non-stationarity. Therefore, the null hypothesis of a unit root cannot be rejected for any variable in level form. In contrast, after first differencing, all variables become strongly significant, and the test statistics are clearly more negative than the critical values at conventional significance levels. This confirms that all variables are integrated of order one, I(1). Importantly, none of the variables are integrated of order two, I(2), thereby satisfying the key econometric requirement for panel cointegration analysis and subsequent long-run estimation techniques. Although CO₂ emissions are measured as a percentage change from a base year (1990), the unit root tests indicate that the variable exhibits I(1) properties within the sample period. This allows the use of panel cointegration estimators such as FMOLS, DOLS, and CCR to investigate long-run relationships among the variables. However, since the dependent variable reflects relative emission dynamics rather than absolute emission levels, the estimated coefficients should be interpreted as long-run associations rather than strict structural elasticities. Finally, the stationarity results provide a solid econometric foundation for the long-run, dynamic, and robustness estimations employed in this study.

4.2.8. Panel Cointegration Test

To examine the existence of long-run equilibrium relationships among the variables, this study employs the Westerlund panel cointegration test, which is particularly suitable in the presence of cross-sectional dependence and slope heterogeneity among EU-25 countries. Unlike conventional first-generation cointegration tests, the Westerlund approach accounts for common factors and cross-sectional interdependence across panel units, thereby providing more reliable and robust cointegration results for highly integrated regional economies such as the European Union. The results reported in

Table 6. Westerlund cointegration test findings.

Test Statistics	Value	p-Value
Gt	−3.742 ***	0.001
Ga	−12.884 ***	0.000
Pt	−8.115 ***	0.000
Pa	−13.507 ***	0.000

Note: *** indicates significance at the 1% level.

provide strong and consistent evidence of panel cointegration among CO₂ emissions, environmental taxation, green innovation, economic growth, imports, exports, and urbanization. Specifically, all four Westerlund test statistics, namely, Gt, Ga, Pt, and Pa, are negative and statistically significant at the 1% significance level, leading to the rejection of the null hypothesis of no cointegration. The significance of both group-mean statistics and panel statistics indicates the existence of stable long-run equilibrium relationships both across individual countries and within the panel as a whole. The Gt and Ga statistics confirm that cointegration exists for at least some cross-sectional units, suggesting that individual EU-25 countries exhibit long-run equilibrium adjustments among the variables. Similarly, the Pt and Pa statistics provide evidence of cointegration for the panel collectively, indicating that the variables move together over time despite short-run fluctuations and temporary deviations from equilibrium. The consistency of the findings across all Westerlund statistics substantially strengthens the reliability and robustness of the cointegration results. The confirmation of long-run cointegration validates the subsequent application of long-run panel estimators, including FMOLS, DOLS, and CCR, for estimating long-run elasticities among the variables. Furthermore, the existence of cointegration also justifies the use of the Panel ARDL framework to examine both short-run dynamics and long-run equilibrium adjustments. In addition, the confirmed long-run relationships provide further support for the implementation of CCEMG and System GMM estimators to address cross-sectional.

5. Results

Table 7. FMOLS, DOLS, and CCR estimation findings.

	FMOLS		DOLS		CCR
	Coefficient	t-Stat	Coefficient	t-Stat	Coefficient	t-Stat
et	2.038428 ***	4.32	4.044049 ***	3.58	1.959646 ***	3.30
gi	−5.676837 **	−2.33	−6.391182 **	−2.19	−5.775968 **	−2.21
gdp	0.5434055 ***	3.29	2.141908 ***	4.63	0.5908418 ***	4.24
import	−0.8920825 ***	−5.93	−2.297153 **	−2.14	−1.020492 ***	−3.78
export	0.5163879 ***	4.49	0.524748 ***	3.26	0.6177219 ***	3.44
urban	0.1227382 ***	3.32	0.0845515 ***	4.21	0.1168046 ***	4.30
_cons	−8.429004 ***	−3.24	−6.032685 ***	−5.15	−7.506553 ***	−5.21

Note: t-statistics are reported alongside coefficient estimates. Statistical significance is evaluated using conventional levels, where, **, and *** denote significance at the 5%, and 1% levels, respectively.

presents the baseline estimations using FMOLS, DOLS, and CCR, providing consistent evidence on the long-run determinants of CO₂ emissions across EU-25 countries over the period 2000–2021. Across all three estimators, the results reveal a stable and robust pattern of relationships.

Across all specifications, environmental taxation exhibits a positive and statistically significant relationship with CO₂ emissions. The estimated elasticities indicate that a 1% increase in environmental taxation is associated with an increase in CO₂ emissions ranging from approximately 1.96% to 4.04%. In the EU-25 context, this result can be interpreted in light of the structural design of environmental tax systems rather than their nominal levels. Many member states rely heavily on energy taxes that are often fiscally motivated rather than strictly environmental in design. In addition, exemptions for energy-intensive and trade-exposed sectors, particularly in Central and Eastern European economies and industrialized Western members, reduce the deterrent effect of taxation on emissions. Furthermore, environmental taxes are frequently introduced as part of broader fiscal consolidation strategies, meaning they may rise in response to higher energy consumption and emissions rather than actively causing them. This reinforces the interpretation of a reactive rather than preventive policy structure across the EU-25.

In contrast, green innovation, proxied by environment-related government R&D expenditure, consistently exhibits a negative and statistically significant effect on CO₂ emissions. The estimated elasticities indicate that a 1% increase in green innovation reduces emissions by approximately 5.67% to 6.39%. In EU-25 economies, this effect reflects the strong role of public-sector innovation systems supported by supranational frameworks such as Horizon Europe, the European Green Deal, and national decarbonization strategies. Countries with stronger research ecosystems-particularly Western and Northern Europe-benefit from faster diffusion of clean technologies, including renewable energy integration, smart grids, energy-efficient industrial processes, and electrified transport systems. These innovation channels reduce dependence on fossil fuels and improve energy efficiency, thereby generating substantial long-run emission reductions. The magnitude of the coefficient suggests that innovation-driven structural change is a more powerful mechanism than price-based instruments alone within the EU policy environment.

Economic growth (GDP) is found to have a positive and statistically significant impact on CO₂ emissions across all estimators, with elasticities ranging from approximately 0.54% to 2.14%. In EU-25 countries, this relationship reflects the continued reliance of economic expansion on energy-intensive sectors such as manufacturing, construction, logistics, and heavy industry, particularly in emerging EU economies in Central and Eastern Europe. Although the EU has made progress toward decoupling growth from emissions through efficiency gains and renewable adoption, the results suggest that absolute decoupling remains incomplete. Growth-induced increases in consumption, mobility, and industrial output continue to outweigh efficiency improvements, especially during periods of economic expansion.

Imports are negatively associated with CO₂ emissions, with elasticities ranging from approximately −0.89% to −2.30%. In the EU-25 context, this may reflect structural substitution effects, where imports replace domestically produced goods that are relatively more carbon-intensive. This is particularly relevant for manufacturing sectors that have partially relocated outside certain EU economies while final consumption remains domestic. As a result, increased imports may reduce domestic production-based emissions, even though global emissions are not necessarily reduced. This outcome is consistent with the EU’s high level of trade integration and participation in global value chains, where production and consumption are geographically separated.

Exports show a positive and statistically significant relationship with CO₂ emissions, with elasticities between approximately 0.52% and 0.62%. This result reflects the scale effect associated with export-driven production in EU economies, particularly in industrial hubs such as Germany, Poland, and Italy. Export expansion increases industrial output, energy use, and transport activity, especially in manufacturing and machinery sectors. Although EU production processes are generally becoming more energy-efficient, export-led growth still generates additional environmental pressure when technological improvements are insufficient to fully offset increased output.

Urbanization has a positive and statistically significant effect on CO₂ emissions, with coefficients ranging from approximately 0.08% to 0.12%. In EU-25 countries, urban areas concentrate economic activity, infrastructure, transportation systems, and residential energy consumption. Cities such as Paris, Berlin, Warsaw, and Madrid act as major consumption and production hubs, where increased population density raises demand for electricity, heating, mobility, and services. Although urbanization can generate efficiency gains through scale economies and public transport systems, these benefits are not yet sufficient to offset rising energy demand in rapidly expanding urban regions, particularly in Eastern and Southern Europe.

Finally, the consistency of results across FMOLS, DOLS, and CCR estimators reinforces the robustness of the findings and highlights the complex, multi-dimensional nature of environmental outcomes in the EU-25. The results suggest that innovation-led policies are significantly more effective in reducing emissions than price-based instruments alone, while economic growth, trade expansion, and urbanization continue to exert upward pressure on emissions. This underscores the need for an integrated policy approach combining environmental taxation reform, stronger innovation systems, trade-adjusted environmental regulation, and sustainable urban development strategies tailored to the heterogeneous structure of EU economies.

Robustness Check by Using Panel ARDL-PMG and System GMM Estimations

Table 8. Robustness check estimation-ARDL PMG.

	(1)	(2)
VARIABLES	__ec	SR
__ec		−0.267 ***
		(0.0609)
D. et		−2.457 ***
		(2.169)
D. gi		−0.661 **
		(0.667)
D. gdp		−0.170 ***
		(0.201)
D. import		0.0897 **
		(0.0588)
D. export		0.0791 ***
		(0.0690)
D. urban		−10.42 ***
		(6.946)
et	2.622 **
	(1.831)
gi	−0.0694 ***
	(0.230)
gdp	3.333 ***
	(0.454)
import	−0.628 ***
	(0.220)
export	0.0988 **
	(0.223)
urban	−4.856 ***
	(0.476)
Constant		86.98 ***
		(22.53)
Observations	483	483

Standard errors in parentheses. *** p < 0.01, ** p < 0.05.

presents the robustness analysis using the Pooled Mean Group (PMG) ARDL estimator, which allows for the simultaneous examination of long-run equilibrium relationships and short-run dynamic adjustments between CO₂ emissions, environmental policy variables, and macroeconomic factors across EU-25 countries over the period 2000–2021.

The error correction term is negative and statistically significant, confirming the existence of a stable long-run cointegrating relationship among the variables. The magnitude of the coefficient suggests a moderate speed of adjustment, implying that approximately 27% of short-term deviations from long-run equilibrium are corrected within one period. This indicates that although EU economies adjust toward equilibrium, structural rigidities, policy heterogeneity, and differences in energy systems slow the convergence process.

In the long-run specification, the PMG results broadly confirm the baseline findings. Environmental taxation remains positive and statistically significant, suggesting that higher environmental tax levels are still associated with increased emissions. In the EU-25 context, this reinforces the interpretation that current environmental tax systems may be insufficiently stringent or unevenly implemented across member states, particularly due to sectoral exemptions and differences in national tax design. As a result, environmental taxation appears to reflect existing emission structures rather than actively driving reductions.

Environment-related R&D expenditure continues to show a negative and statistically significant effect, confirming its central role in promoting long-run emission reductions. In EU economies, this result reflects the importance of publicly supported innovation systems, including EU-wide programs such as Horizon Europe and national green transition funds. These investments enhance technological capability, support the diffusion of low-carbon technologies, and improve energy efficiency across industrial and transport sectors, leading to sustained reductions in carbon intensity.

Economic growth retains a positive long-run relationship with CO₂ emissions, indicating that expansion in output continues to exert environmental pressure in EU countries. Despite ongoing decarbonization efforts, economic activity in sectors such as manufacturing, construction, and transport remains energy-intensive, particularly in emerging EU economies where industrial restructuring is still ongoing.

Trade variables also remain consistent with the baseline results. Imports are negatively associated with emissions, suggesting that increased import penetration may reduce domestic production of relatively carbon-intensive goods, reflecting structural substitution within EU production systems. Exports, however, maintain a positive long-run effect, indicating that export-led production expansion continues to increase energy demand and emissions, especially in manufacturing-oriented economies integrated into global value chains.

A notable divergence from the baseline results emerges in the case of urbanization. Unlike previous estimations, urbanization exhibits a negative and statistically significant long-run effect on CO₂ emissions. This suggests that, over time, urban development in EU-25 countries may contribute to environmental improvement through efficiency gains. These gains likely arise from dense urban infrastructure, improved public transportation systems, district heating networks, and the adoption of smart city technologies, which reduce per capita energy consumption and improve resource efficiency.

The short-run estimates reveal more volatile and immediate responses to shocks. Environmental taxation and environment-related R&D expenditure both exhibit negative short-run effects on emissions, suggesting that policy interventions can have immediate mitigating impacts, even if long-run effectiveness varies. This indicates that fiscal and innovation policies may influence behavioral adjustments and energy use decisions in the short term before structural effects fully materialize.

Trade variables display positive short-run effects, reflecting the immediate environmental costs associated with increased production and trade activity. Short-term fluctuations in exports and imports are closely linked to industrial output and logistics activity, which temporarily increase energy demand and emissions.

Economic growth and urbanization show negative short-run coefficients, suggesting that short-term adjustments may involve cyclical efficiency improvements, temporary demand shifts, or structural reallocation effects that reduce emissions in the short term. These effects, however, differ from long-run trends and highlight the importance of distinguishing between transitory dynamics and structural transformations.

Taken together, the PMG results reinforce the robustness of the baseline estimations while adding important dynamic insights. The evidence confirms that environmental outcomes in EU-25 countries are shaped by both long-run structural forces and short-run adjustment mechanisms. While innovation consistently emerges as a key driver of emission reductions, environmental taxation alone appears insufficient without stronger design and coordination. Trade and growth continue to exert upward pressure on emissions, while urbanization demonstrates context-dependent effects across time horizons. These findings underscore the importance of a balanced and multi-dimensional policy framework that integrates fiscal instruments, innovation systems, trade regulation, and urban sustainability strategies to achieve long-term environmental objectives in the European Union.

To further strengthen the robustness and reliability of the empirical analysis, this study additionally employs the System Generalized Method of Moments and the Common Correlated Effects Mean Group estimators. The empirical outcomes of these techniques are provided in

Table 9. Robustness check estimation-System GMM and CCEMG.

Variables	Coefficient	CCEMG
et	1.684215 ***	1.842517 ***
gi	−4.928316 **	−5.214883 **
gdp	0.774518 ***	0.693418 ***
import	−0.948621 ***	−0.884276 ***
export	0.571440 ***	0.548117 ***
urban	0.109372 ***	0.102736 ***
L. CO2	0.462185 ***	-
_cons	−6.918542 ***	−7.104882 ***
AR(1) p-value	0.021
AR(2) p-value	0.287
Hansen J-test p-value	0.412
Number of Instruments	24
Number of Groups	25
Observations	550

Standard errors in parentheses. *** p < 0.01, ** p < 0.05.

. The use of these advanced panel econometric techniques is particularly important given the potential presence of endogeneity, reverse causality, dynamic persistence, cross-sectional dependence, and slope heterogeneity among EU-25 countries over the period 2000–2021.

The System Generalized Method of Moments estimator was applied to address possible endogeneity problems arising from simultaneity, omitted variable bias, and the bidirectional relationship between environmental degradation and the explanatory variables. Environmental taxation may respond to rising pollution levels, renewable energy investment may increase following environmental deterioration, and economic growth and trade activities may simultaneously influence and be influenced by CO₂ emissions. Furthermore, environmental degradation tends to exhibit dynamic persistence because current emission levels are strongly affected by previous environmental conditions due to structural dependence in industrial production systems, fossil fuel consumption, transportation infrastructure, and energy markets.

To control for these issues, lagged levels and lagged first differences of the endogenous variables were employed as internal instruments. Specifically, lagged values of environmental taxation, green innovation, economic growth, imports, exports, urbanization, and the lagged dependent variable were used as instruments because they are strongly correlated with current values while remaining uncorrelated with the contemporaneous error term. The use of internal instruments is particularly suitable in macro-panel studies where theoretically valid external instruments are difficult to obtain.

The System GMM results reveal that environmental taxation positively and significantly influences CO₂ emissions. The estimated coefficient indicates that a one-unit increase in environmental taxation is associated with an approximately 1.68-unit increase in CO₂ emissions, holding other factors constant. This positive relationship suggests that existing environmental tax policies within EU-25 countries may not yet be sufficiently stringent or effective in reducing pollution-intensive activities. This outcome may reflect adjustment costs, incomplete environmental tax coverage, or the continued dominance of carbon-intensive industrial sectors despite ongoing environmental fiscal reforms.

Green innovation shows a negative and statistically significant relationship with environmental degradation. The results indicate that higher levels of green innovation are associated with lower CO₂ emissions, highlighting its role in supporting environmentally sustainable investment and accelerating the transition toward cleaner economic activities. This finding is consistent with Ecological Modernization Theory, which emphasizes that innovation, institutional support, and public investment are key drivers of technological upgrading and structural transformation, ultimately contributing to the reduction of environmental pressure.

Economic growth exerts a positive effect on CO₂ emissions, as a 1% increase in GDP increases environmental degradation by approximately 0.77%. This finding suggests that economic expansion within EU-25 countries continues to generate environmental pressure through higher industrial output, transportation activities, and energy demand. Imports are negatively associated with CO₂ emissions. The estimated coefficient indicates that a one-unit increase in imports is associated with an approximately 0.95-unit decrease in CO₂ emissions, holding other factors constant. This finding suggests that greater import penetration may facilitate access to cleaner technologies, environmentally efficient intermediate goods, or less carbon-intensive production processes. Conversely, exports positively influence environmental degradation, as a 1% increase in exports increases CO₂ emissions by approximately 0.57%, reflecting the environmental costs associated with export-oriented industrial production and transportation activities.

Urbanization exerts a positive effect on environmental degradation. The estimated coefficient indicates that a one-unit increase in urbanization is associated with an approximately 0.11-unit increase in CO₂ emissions. This finding reflects the environmental consequences of rising urban populations, increasing infrastructure expansion, transportation intensity, and higher energy consumption within urban areas.

The coefficient of the lagged dependent variable confirms the existence of dynamic persistence in environmental degradation. Specifically, a 1% increase in previous CO₂ emissions leads to an approximately 0.46% increase in current CO₂ emissions, indicating that past environmental conditions strongly influence present emission levels.

The diagnostic tests further confirm the validity and robustness of the System GMM estimation. The AR(1) test indicates the expected presence of first-order serial correlation in differenced residuals, while the AR(2) test confirms the absence of second-order serial correlation, satisfying a key requirement of dynamic panel estimation. Moreover, the Hansen J-test confirms the validity and exogeneity of the selected instruments. The number of instruments also remains below the number of cross-sectional groups, thereby minimizing the risk of instrument proliferation and overfitting.

To additionally account for cross-sectional dependence and slope heterogeneity among EU-25 countries, the Common Correlated Effects Mean Group estimator developed by Pesaran was employed as a robustness estimation technique. The CCEMG estimator is particularly appropriate in the context of EU economies because member countries are highly interconnected through common environmental regulations, energy markets, trade integration, monetary coordination, and EU-wide climate policies. Conventional panel estimators may therefore produce biased and inconsistent estimates if cross-sectional dependence is ignored.

The CCEMG results remain largely consistent with the findings obtained from FMOLS, DOLS, CCR, Panel ARDL, and System GMM estimations, thereby reinforcing the robustness and stability of the empirical analysis. Environmental taxation continues to exert a positive and statistically significant effect on CO₂ emissions, while green innovation consistently reduces environmental degradation. Similarly, economic growth, exports, and urbanization maintain positive relationships with CO₂ emissions, whereas imports continue to exhibit a negative association with environmental degradation. The consistency of coefficient signs and statistical significance across multiple econometric approaches confirms that the estimated long-run relationships are stable and not sensitive to alternative estimation techniques.

Taken together, the combined use of System GMM and CCEMG substantially strengthens the econometric rigor of this study by simultaneously addressing endogeneity, dynamic persistence, cross-sectional dependence, and heterogeneity across EU-25 countries. The findings therefore provide more reliable and policy-relevant evidence regarding the determinants of CO₂ emissions and the role of green innovation and environmental policy instruments in promoting environmental sustainability within the European Union.

6. Discussion

A key finding of this study is the positive and statistically significant impact of environmental taxation on CO₂ emissions across all long-run estimators. At first sight, this appears counterintuitive since environmental taxation is theoretically designed to internalize externalities and reduce pollution by increasing the relative cost of carbon-intensive activities. However, within the EU-25 context, this outcome is more plausibly interpreted through structural and institutional characteristics of environmental fiscal systems rather than a direct causal increase in emissions. Environmental tax revenues tend to be higher in economies where emissions are already elevated due to historically large industrial bases, energy-intensive production structures, and transport demand. As a result, taxation is often reactive, meaning it is implemented in response to existing environmental pressures rather than functioning as a preventive mechanism. This generates a positive correlation between taxation levels and emissions in aggregated long-run estimations. In addition, cross-country differences in industrial structure within the EU-25 imply that countries with stronger manufacturing bases and higher fossil fuel dependence naturally generate both higher emissions and higher environmental tax revenues, reinforcing this co-movement in the data.

In addition, the effectiveness of environmental taxation varies significantly across EU member states due to differences in tax design, sectoral coverage, and enforcement capacity. Many countries apply reduced rates, exemptions, or compensatory mechanisms for energy-intensive and trade-exposed industries to protect competitiveness. While such measures are politically and economically important, they reduce the marginal incentive to shift away from fossil fuel dependence and slow the structural transition toward cleaner production technologies. This is particularly relevant in integrated markets such as the EU, where policy coordination is partial and national fiscal autonomy still plays a major role in environmental taxation outcomes. This is consistent with evidence suggesting that environmental taxes only become strongly emission-reducing when they reach sufficiently high levels and are embedded within broader regulatory frameworks (Mehta & Prajapati, 2024; Murad et al., 2025). It also aligns with studies indicating that poorly structured or low-intensity environmental taxation may coexist with rising emissions, particularly in heterogeneous policy environments such as the EU (Murad et al., 2025; M. Wang & Kuusi, 2024). Furthermore, the observed positive relationship may also reflect a transitional phase in EU climate policy where taxation is still evolving and has not yet fully aligned with sector-specific decarbonization targets.

The Panel ARDL results provide additional nuance by showing that environmental taxes exert a negative effect in the short run, suggesting that behavioral responses and energy use adjustments may initially occur following policy changes. Firms and households may reduce consumption, improve efficiency, or adopt cleaner inputs in response to higher costs. However, these short-run adjustments may not fully translate into long-run structural change due to capital stock rigidity, long investment cycles, and the persistence of carbon-intensive infrastructure. Over time, the initial behavioral response may be offset by structural lock-in effects, where existing industrial capacity limits the extent of deep decarbonization. This dynamic interpretation is consistent with evidence that environmental fiscal instruments can induce gradual behavioral change over time rather than immediate structural transformation (Mehta & Prajapati, 2024; Murad et al., 2025). It also suggests that environmental taxation is more effective when combined with complementary policies such as innovation subsidies, regulatory standards, and carbon market mechanisms that jointly address both price and non-price barriers.

In contrast, green innovation proxied by environment-related government R&D expenditure exhibits a consistently negative and statistically significant effect on CO₂ emissions across all estimation techniques. This finding is highly robust and suggests that public investment in environmental innovation plays a central role in reducing emissions through technological upgrading and structural transformation. Unlike price-based instruments, green innovation operates through the supply side of the economy by improving production technologies, enhancing energy efficiency, and enabling the development and diffusion of low-carbon solutions. This makes its impact more persistent and structurally embedded compared to fiscal measures. It also implies that innovation effects accumulate over time, producing long-term changes in production structure rather than short-term behavioral adjustments.

Within the EU context, this result reflects the importance of coordinated innovation strategies such as Horizon Europe and national green transition programs, which support research, knowledge creation, and diffusion of clean technologies. These institutional frameworks reduce technological uncertainty, promote cross-country learning, and facilitate convergence toward cleaner production standards across member states. These investments facilitate long-term decarbonization by reducing the carbon intensity of industrial production, transport systems, and energy use. This is consistent with empirical evidence showing that green innovation, environmental R&D, and innovation-driven investment reduce CO₂ emissions across both developed and emerging economies (Zhang et al., 2022; N. Amin et al., 2025; Wu et al., 2021; Umar & Safi, 2023). The robustness of this result across all estimators suggests that innovation is a more structurally stable driver of emission reduction than fiscal instruments alone, particularly in heterogeneous regional systems such as the EU-25.

Economic growth is found to have a positive and statistically significant impact on CO₂ emissions, consistent with the early stages of the Environmental Kuznets Curve hypothesis. This indicates that economic expansion in EU-25 countries continues to generate environmental pressure through higher production, consumption, and energy demand. Despite progress in decoupling emissions from GDP growth, this decoupling remains partial and uneven across member states. Energy-intensive sectors such as manufacturing, construction, logistics, and transport still dominate in several economies, particularly in lower-income or transitioning EU countries. In these contexts, growth is still closely tied to fossil fuel-based energy systems, limiting the immediate environmental benefits of economic expansion. The variation in coefficient magnitude across estimators suggests heterogeneity in growth-emission relationships, reflecting differences in industrial structure, energy mix, technological readiness, and policy effectiveness. The Panel ARDL results further indicate a negative short-run effect of GDP growth, suggesting temporary efficiency gains or cyclical adjustments that do not persist in the long run. This may reflect short-term productivity improvements, utilization of idle capacity, or temporary shifts in production composition during economic fluctuations (Aye & Edoja, 2017; Mohammed et al., 2023; Balsalobre-Lorente et al., 2018).

The disaggregated trade variables reveal asymmetric effects on emissions, highlighting the complex role of international trade in shaping environmental outcomes. Imports are consistently associated with lower CO₂ emissions in the long run, suggesting that EU countries may be outsourcing carbon-intensive production to external economies. This can be interpreted as a production substitution effect, where domestic production of emissions-intensive goods declines while consumption demand is increasingly met through imports. In theoretical terms, this mechanism is consistent with ecological interdependence and global value chain reallocation, where production is spatially fragmented according to comparative advantage and environmental regulation differentials. However, the empirical specification does not explicitly model embodied emissions, sectoral trade composition, or consumption-based accounting, and therefore cannot directly identify carbon leakage channels. As a result, the findings should be interpreted as reduced-form evidence that is suggestive of leakage-type dynamics rather than definitive proof. This limitation points to the need for future research incorporating input-output tables or consumption-based CO₂ accounting to more precisely capture cross-border emission transfers. This interpretation is broadly consistent with empirical literature on embodied emissions in trade and carbon leakage concerns (Sufyanullah et al., 2022; Q. Wang et al., 2023; M. Wang & Kuusi, 2024) and aligns with consumption-based perspectives in environmental economics that emphasize emissions displacement rather than elimination.

Exports, in contrast, are found to increase emissions, reflecting higher domestic production associated with export-oriented industries. Theoretically, this result can be linked to the scale effect within the Environmental Kuznets Curve (EKC) framework, where increased economic activity initially raises environmental pressure before structural and technological effects dominate at higher income levels. This effect is particularly relevant in manufacturing-intensive EU economies integrated into global value chains, where export expansion increases energy use, transportation activity, and industrial output. At the same time, in the absence of interaction terms capturing technology intensity, environmental regulation heterogeneity, or sectoral composition effects, the estimates should be interpreted as net equilibrium outcomes rather than structural causal parameters. This suggests that the observed relationship reflects the combined influence of production scale and partially offsetting efficiency improvements rather than isolated trade channels. Similar findings are reported in studies showing that export expansion is often associated with higher territorial emissions due to production scale effects (Muhammad et al., 2020; Q. Wang et al., 2023; Balsalobre-Lorente et al., 2018).

Urbanization is found to have a positive and statistically significant impact on emissions in the long run, indicating that rising urban population shares increase environmental pressure through higher energy consumption, mobility demand, and infrastructure expansion. In EU countries, urban areas concentrate economic activity, residential energy use, transport networks, and service industries, all of which contribute to emissions. The long-run positive relationship suggests that urban expansion still relies on energy-intensive systems in many member states, particularly in rapidly growing metropolitan regions. However, the Panel ARDL results show a negative short-run effect, suggesting that urban systems may temporarily improve efficiency through public transport usage, congestion management, and short-term behavioral adjustments. This indicates that urbanization has both efficiency-enhancing and pressure-increasing effects depending on the time horizon considered, and that the net effect depends on the maturity of urban infrastructure and planning systems (Ali et al., 2019; Sethi et al., 2023; Sikder et al., 2022).

Importantly, the findings reveal significant heterogeneity across countries, reflecting differences in income levels, industrial structures, institutional quality, and policy enforcement capacity within the EU-25. High-income countries tend to have stronger environmental institutions, more advanced technologies, and greater capacity to implement effective climate policies, while lower-income or transition economies may still rely on carbon-intensive production structures. This heterogeneity implies that average panel estimates may mask important country-specific dynamics and that policy effectiveness is not uniform across the union. The variation in results across estimators further confirms that environmental relationships are not uniform and depend on structural conditions and stages of development. This is consistent with comparative evidence showing income-specific and institutionally conditioned environmental outcomes (N. Amin et al., 2025; Wu et al., 2021; Guliyev & Seyfullayev, 2025; Umar & Safi, 2023).

To summarize, the consistency of results across FMOLS, DOLS, CCR, and Panel ARDL models strengthens the credibility of the empirical findings and confirms that results are not driven by a single estimation strategy. The use of multiple econometric techniques improves reliability by addressing endogeneity, heterogeneity, and cross-sectional dependence, as emphasized in recent environmental economics literature (Aye & Edoja, 2017; Mohammed et al., 2023; Balsalobre-Lorente et al., 2018; Sikder et al., 2022). Overall, this study provides strong evidence that environmental outcomes in EU-25 countries are shaped by the interaction of fiscal instruments, innovation systems, trade structures, and urban development, with green innovation emerging as the most consistently effective mechanism for long-term emission reduction.

7. Policy Recommendations

The empirical findings of this study generate several policy-relevant implications for achieving sustainable economic growth and environmental improvement within the European Union. Given the observed long-run heterogeneity across EU-25 countries, the results indicate that a uniform policy framework is unlikely to be effective. Instead, policy design should explicitly account for differences in income levels, institutional capacity, and structural economic characteristics.

First, environmental taxation should be redesigned rather than simply expanded. The results suggest that environmental taxes, in their current form, are not sufficiently effective in reducing emissions in the long run, implying that policy effectiveness depends on design features rather than tax presence alone. More specifically, effectiveness can be improved through differentiated and progressively increasing tax schedules linked to sectoral emission intensity, as well as the gradual removal of exemptions for energy-intensive industries. In addition, introducing rule-based adjustment mechanisms-where tax rates increase when emissions exceed predefined thresholds-could strengthen the dynamic responsiveness of fiscal instruments. Earmarking a defined share of environmental tax revenues for green R&D, renewable energy deployment, and energy efficiency upgrades would further enhance the transmission from taxation to decarbonization outcomes.

Second, green innovation should be strengthened through more targeted and outcome-oriented policy instruments. The results indicate that green innovation plays a central role in reducing emissions, but its effectiveness depends on the efficiency and targeting of public investment. Policy design should therefore prioritize mission-oriented R&D programs with clearly defined emission-reduction objectives, alongside competitive funding mechanisms that reward high-impact innovations in renewable energy, storage technologies, and industrial decarbonization. Strengthening collaboration between public research institutions and industry is also essential to accelerate technology diffusion and improve commercialization outcomes. In addition, establishing harmonized EU-wide evaluation frameworks for environmental R&D would help ensure accountability and allow for performance-based allocation of innovation funding.

Third, decoupling economic growth from emissions requires sector-specific rather than aggregate policy interventions. The positive association between GDP and emissions suggests that growth remains partially dependent on carbon-intensive production structures. Accordingly, policy should focus on differentiated carbon pricing mechanisms across sectors, stricter efficiency standards in manufacturing and transport, and targeted subsidies for low-carbon technologies such as electrification and hydrogen-based systems. This sectoral approach is essential for addressing structural emissions rather than relying on economy-wide averages.

Fourth, trade policy should be more explicitly linked to production-based emissions responsibility. The asymmetric effects of exports and imports suggest that trade influences emissions through production relocation and scale effects. In this context, instruments such as the Carbon Border Adjustment Mechanism (CBAM), product-level carbon footprint requirements, and enhanced supply-chain transparency can help align trade flows with EU climate objectives. These measures can reduce the risk of carbon leakage while ensuring that external trade does not undermine domestic decarbonization efforts.

Fifth, urbanization policy should prioritize qualitative transformation rather than quantitative expansion. The results imply that the environmental impact of urbanization depends critically on infrastructure quality and planning efficiency. Policy should therefore focus on investments in low-carbon urban infrastructure, including electrified public transport systems, energy-efficient buildings, and integrated district heating and cooling networks. Improved spatial planning and densification strategies can further reduce transport demand and enhance energy efficiency in urban systems.

Finally, the heterogeneity across countries suggests that policy effectiveness is conditioned by income levels and institutional development. Lower-income EU member states should prioritize institutional strengthening, access to green finance, and technology transfer mechanisms, while higher-income countries should focus on innovation-led decarbonization and stricter regulatory enforcement in high-emission sectors. At the EU level, stronger coordination through cohesion policy instruments, structural funds, and the European Green Deal is necessary to reduce fragmentation and ensure convergence in environmental policy implementation.

In conclusion, achieving environmental sustainability in the European Union requires not only an integrated policy framework but also precisely designed instruments with clear transmission channels, measurable targets, and sector-specific implementation strategies. The effectiveness of these policies depends critically on institutional capacity, structural economic characteristics, and cross-country heterogeneity, underscoring the need for a differentiated yet coordinated EU-wide policy architecture.

8. Conclusions

This study investigates the determinants of CO₂ emissions in EU-25 countries over the period 2000–2021, with particular attention to environmental taxation and green innovation alongside key macroeconomic and structural variables. A comprehensive panel econometric framework is employed, incorporating FMOLS, DOLS, CCR, Panel ARDL, CCEMG, and System GMM estimators to ensure robust and consistent inference across both long-run and short-run dynamics.

The empirical results provide several important insights. Environmental taxation is found to have a positive and statistically significant association with CO₂ emissions in the long run, suggesting that its effectiveness depends strongly on policy design, coverage, and enforcement. This outcome indicates that environmental taxation in EU countries may still be largely reactive or unevenly implemented, and therefore not yet sufficiently effective in driving emission reductions at the aggregate level.

In contrast, green innovation proxied by environment-related government R&D expenditure consistently shows a negative and statistically significant effect on CO₂ emissions. This highlights the crucial role of public investment in environmental innovation and technological development in supporting long-term decarbonization and structural transformation. Among the policy instruments considered, green innovation emerges as a more stable and effective channel for reducing emissions.

Economic growth is positively associated with CO₂ emissions, indicating that economic expansion continues to exert environmental pressure despite ongoing sustainability efforts. Trade effects are asymmetric, with exports increasing emissions through higher production activity, while imports reduce emissions, likely reflecting the relocation of carbon-intensive production outside domestic economies. Urbanization also contributes to higher emissions in the long run, although short-run adjustments suggest temporary efficiency gains and structural adaptation.

Finally, the consistency of results across all estimation techniques confirms the robustness of the findings. The evidence suggests that CO₂ emissions in EU-25 countries are shaped by a combination of fiscal policy, innovation capacity, trade structure, and urban development. The results emphasize that achieving long-term environmental sustainability requires a balanced policy mix, where environmental taxation is strengthened and complemented by stronger investment in green innovation and structural transformation policies.

Despite its contributions, this study has several limitations that should be acknowledged. First, the proxy for green innovation-environment-related government R&D expenditure captures only the public component of environmental innovation and does not reflect broader market-based green finance instruments such as green bonds, green credit, or private sustainable investment flows. Second, the analysis is based on aggregate macro-panel data and does not account for sectoral heterogeneity or differences in carbon intensity across production structures; therefore, embodied carbon flows and carbon leakage mechanisms cannot be directly identified. Third, although advanced panel econometric methods (System GMM and CCEMG) are applied, potential endogeneity, unobserved heterogeneity, and heterogeneous policy transmission channels cannot be fully ruled out.

Finally, some relevant determinants-such as energy structure, fossil fuel dependence, industrial composition, and EU ETS participation-are not included due to data and scope constraints. Future research could address these limitations by incorporating richer measures of green finance, sectoral and embodied carbon data, and nonlinear or interaction-based econometric frameworks, as well as policy-specific identification strategies to better capture causal environmental effects.