Energy–Growth Nexus in European Union Countries During the Green Transition

Abstract

This study investigates the relationship between economic growth and energy consumption—both renewable and non-renewable—in European Union countries during the green transition. Using a panel dataset of 28 EU countries from 1995 to 2021, we employ econometric techniques—including the Westerlund cointegration test and a fixed-effect panel threshold model—to assess long-term equilibrium relationships. The results indicate that while both renewable and non-renewable energy consumption are associated with economic growth, their roles differ. Renewable energy consumption shows a positive but less robust relationship with economic growth. In contrast, non-renewable energy consumption demonstrates a more robust bidirectional causality with economic growth, indicating a more intertwined relationship with economic growth during the study period. Interestingly, in countries with high levels of non-renewable energy consumption—classified as regime 2 in the panel threshold model—increased non-renewable energy consumption is associated with a decrease in economic activity. Our results have significant policy recommendations, indicating that promoting renewable energy sources does not hinder economic growth. Moreover, such promotion has the potential to contribute substantially to economic growth in the future. Therefore, in addition to other crucial benefits, such as increased energy security, the development of renewable energy sources does not threaten the economy. This is particularly relevant as many EU countries, including Poland, Romania, Hungary, Bulgaria, Slovakia, and Lithuania, still have underdeveloped renewable energy sectors.

1. Introduction

Introducing the definition of the green transition requires a brief introduction to the plan that initiated transformations in Europe, dating back to 2019. At that time, the European Commission announced the European Green Deal, referring to a plan to (re)build the EU’s sustainable economy by addressing climate and environmental challenges across all policy areas [1]. The main goal of the European Green Deal is to achieve climate neutrality by 2050, making Europe the first climate-neutral continent. The planned actions include investments in clean technologies, a fair transition for all sectors and regions, and support for sustainable trade [1]. The implementation of the European Green Deal comes with numerous challenges, such as the transition to renewable energy sources, improving energy efficiency, increasing the share of energy from renewable sources, and reducing dependence on fossil fuels. These goals aim not only to achieve climate neutrality, but also to gain a competitive edge in the global market and redefine economic growth, taking into account environmental and social aspects [2,3].

The “green transition”, in turn, has a broader scope, as it can be defined as a global process of change that involves not only initiatives undertaken by public institutions, but also actions taken by the private sector, civil society, and local governments [4,5]. It is a long-term adaptive process aimed at adjusting to changing environmental and economic conditions, as well as minimizing the negative impact of human activity on the natural environment. Through the development of the green transition, planned actions include the expansion of renewable energy, the improvement of the energy efficiency, biodiversity protection, and the promotion of sustainable practices in agriculture and industry [6]. This process continues to evolve, and different countries and communities are introducing unique strategies tailored to their specific conditions and challenges, covering an increasing number of aspects. Green growth emphasizes, above all, the greening and intensification of the economy [7,8].

The shift to a green economy during economic development involves a transition toward a more sustainable and environmentally neutral economic model. Recently, there has been a growing interest in eco-friendly products and services, as well as sustainable business practices, which has prompted companies to adjust to consumer expectations [9]. The comprehensive improvement of green total factor productivity is a fundamental way to promote the green transition in the development process and achieve high-quality economic growth. An important component of the green transition is the energy transition, which covers a wide range of aspects, such as the development of innovative renewable technologies, improving the energy efficiency, modernizing the energy infrastructure, analyzing the energy policy, and promoting ecological awareness and changes in social behavior through education in this field [10,11].

The European Union has set specific targets regarding the reduction of greenhouse gas emissions and the increase in the share of renewable energy in total energy production. These goals impact various economic sectors and require actions to improve the energy efficiency and promote clean technologies. It is widely recognized that the green transition is essential for achieving sustainable development, as it can lead to economic growth while simultaneously ensuring environmental and social well-being [12,13].

The green transition creates incentives to accelerate the decarbonization of the economy and shift to more sustainable energy management models. At the same time, policies supporting education and building public awareness play an important role in accepting change and supporting energy efficiency measures. Adapting these measures to the specifics of individual EU member states allows the effective implementation of climate goals and new prospects for economic development. With a green transition, the EU energy sector can become more competitive, resilient to crises, and better prepared for future challenges.

This study aims to investigate the relationship between economic growth and energy consumption, focusing on both renewable and non-renewable sources, within the European Union. By analyzing a panel dataset covering 28 EU countries (including the United Kingdom) over the period from 1995 to 2021, we seek to provide a long-term perspective on this relationship during a critical phase of environmental policy evolution in Europe.

Grounded in the neoclassical production function, our theoretical framework incorporates gross fixed capital formation and the total labor force as fundamental determinants of economic growth. We began our analysis by testing for cross-sectional dependence among the variables using Pesaran’s CD test, followed by the Pesaran and Yamagata Delta test to examine slope homogeneity. To check for stationarity in our panel data, we applied the cross-sectional Augmented Dickey–Fuller test and the cross-sectional Im, Pesaran, and Shin test. We then evaluated cointegration using the Westerlund cointegration test. To accommodate heterogeneous slope coefficients, we employed various econometric techniques such as the Mean Group estimator, the Augmented Mean Group estimator, and Fully Modified Ordinary Least Squares to capture the dynamic relationships within the data. Additionally, we estimate a fixed-effect panel threshold model, which allows the coefficient of non-renewable energy consumption to differ depending on whether the threshold variable—non-renewable energy consumption—is below or above the estimated threshold. Finally, we conducted the Granger causality test by Dumitrescu and Hurlin to determine whether one series could predict or influence another. These approaches allow us to capture the dynamic relationships within the data accurately.

By focusing on the EU during its green transition, this study fills a gap in the existing literature by providing a long-term perspective on the impact of renewable and non-renewable energy consumption on economic growth within this specific context. Specifically, it provides empirical evidence of the non-linear relationship between non-renewable energy consumption and economic growth. The findings contribute to the policy discourse on sustainable development, highlighting the nuanced roles of different energy types in promoting long-term economic growth while transitioning toward a more sustainable and environmentally friendly energy system in the European Union. Understanding this relationship is essential for EU policymakers to develop strategies that support the shift to renewable energy while maintaining economic prosperity, thereby advancing sustainable development goals.

The remainder of this paper is organized as follows: Section 2 reviews the key literature on the energy–growth nexus, emphasizing studies relevant to the EU context. Section 3 describes the data sources and outlines the econometric methods used in the analysis. Section 4 presents the empirical results and compares them with previous research findings. Finally, Section 5 concludes with policy recommendations and suggests avenues for future research.

2. Literature Review

Advancements in renewable energy research can contribute to sustainable economic growth while simultaneously supporting the green transition. Therefore, technological innovations and supportive policies are crucial in accelerating the achievement of ecological and economic goals [14]. Given global climate change and diminishing fossil fuel reserves, the nexus between renewable energy consumption and economic growth has become a focal point for policymakers and scholars [15,16,17,18,19,20]. Understanding this relationship is essential for developing strategies that promote sustainable economic development while mitigating environmental degradation.

Most studies have focused on the link between energy consumption and income or the energy–income–emissions nexus. Since economic growth is closely tied to energy availability and consumption, which drives economic and industrial activities, increasing attention has also been devoted to exploring the impact of renewable energy consumption on economic growth. These studies examine its effects on aspects such as energy security, technological innovation, investment, and research and development expenditures [21,22,23,24,25].

The utilization of different energy sources, both renewable and non-renewable, constitutes a pivotal factor in fostering economic growth. Analyzing the causal linkages between energy consumption and economic expansion is of paramount importance, as a stable and robustly growing economy can generate the requisite financial resources to support the advancement and efficient deployment of energy systems [26,27,28]. Attention should also be given to the four distinct hypotheses regarding the relationship between energy consumption and economic growth. Shahbaz et al. [29] synthesized the main findings from the literature, highlighting that these four hypotheses have been confirmed. The first, known as the “non-causality hypothesis” or “neutrality hypothesis”, suggests no statistically significant relationship between energy consumption and the production of final goods in the economy. The confirmation of this hypothesis indicates that policies aimed at reducing energy consumption to lower greenhouse gas emissions would not negatively impact domestic production. This relationship has been observed by, among others, Mbanda et al. [30] and Zeren et al. [31].

The second hypothesis is the “uni-directional causality from economic growth” or “conservation hypothesis”, which asserts that real GDP growth affects energy consumption. In this case, decisions to reduce energy consumption would have only a marginal impact on economic dynamics. Examples of such analyses can be found in the works of Sadorsky [32]. The third hypothesis, “uni-directional causality from energy consumption” or the “growth hypothesis”, assumes that energy consumption influences economic growth. If there is a positive relationship between these variables, then measures aimed at reducing pollution may negatively impact domestic production. Examples of such analyses can be found in the works of Bhuiyan et al. [33], Apergis and Payne [34], Pao and Fu [35], Dergiades, Martinopoulos, and Tsoulfidis [36], as well as Fuinhas and Marques [37].

The fourth hypothesis is the “bi-directional causality” or “feedback hypothesis”, which suggests that energy consumption and economic growth are interdependent. Increased energy consumption leads to higher real GDP, which in turn positively affects further energy consumption in the country. This relationship has been observed by, among others, Fuinhas and Marques [36], Ozturk and Bilgili [37], as well as Apergis and Payne [34], and Ahmed and Shimada [38].

Recent studies conducted in European Union countries merit particular attention. The research by Tutak and Brodny [39] showed that the increase in renewable energy consumption between 2000 and 2019 had a positive impact on economic growth in European Union countries, with the growth dynamics being stronger in the “old” EU countries (EU-14) than in the “new” ones (EU-13). Countries that increased the share of renewable energy in their energy mix experienced higher economic growth due to lower energy costs, the creation of new jobs, technological innovations, and enhanced energy security. Additionally, the rise in renewable energy consumption contributed to a significant reduction in greenhouse gas emissions. It decreased the use of conventional energy sources, which improved the environmental quality and reduced the dependence on fossil fuel imports. These results confirm a strong, positive correlation between the development of renewable energy sources and economic growth, highlighting the need for the further support of policies promoting sustainable development in EU countries.

The bi-directional relationship between economic growth and renewable energy consumption in EU countries was analyzed by Radmehr, Henneberry, and Shayanmehr [40]. The research results indicate that renewable energy consumption supports economic growth and that the development of the renewable energy sector in one country can positively affect neighboring countries. Bhattacharya [20] emphasized that higher incomes may lead to increased energy consumption and higher greenhouse gas emissions, particularly in developing countries. Energy policies, investments in renewable energy, and appropriate regulations can significantly reduce emissions and support economic growth. Different perspectives have emerged in studies on different regions exploring the relationship between renewable energy consumption, economic growth, and their impact on greenhouse gas emissions. Acheampong, Dzator, and Savage [41] investigated the causal link between renewable energy, economic growth, and carbon dioxide emissions in 45 Sub-Saharan African countries from 1960 to 2017. Utilizing the Generalized Method of Moments–Panel Vector Autoregression (GMM-PVAR) method, their findings revealed a bi-directional causal relationship between economic growth and renewable energy consumption. Similarly, Koengkan, et al. [42] employed a PVAR model in 12 Latin American countries, confirming that renewable energy consumption promotes economic growth.

In OPEC countries, Keshavarzian and Tabatabaienasab [43] examined the relationship between renewable and non-renewable energy consumption and economic growth from 1980 to 2018. Their findings confirmed the neutrality hypothesis for Angola, Iraq, Nigeria, Venezuela, and Congo. Similarly, Göksu [44] explored the asymmetric effects of fluctuations in non-renewable energy consumption on economic growth in Turkey. This study underscores the complexity of the relationship between different types of energy consumption and economic growth, suggesting that the impact of renewable energy on the economic performance may be less clear in certain regions, thereby aligning with Menegaki’s [45] neutrality hypothesis.

Table 1. Overview of the primary studies in the area.

Authors	Research Sample/Period	Causality Results
Al-Mulali [46]	30 major nuclear-energy-consuming countries (1990–2010)	FEC <=> GDP GDP => FEC NEC => GDP
Nasreen and Anwar [47]	15 Asian countries (2000–2010)	GDP <=> EC GDP => EC
Marques and Fuinhas [48]	24 European countries (1990–2007)	REC <=> GDP
Dedeoğlu and Kaya [49]	OECD countries (1990–2011)	EC <=> GDP
Ciarreta and Zarraga [50]	12 European countries (1970–2007)	EL => GDP
Mozumder and Marathe [51]	Bangladesh (1980–2008)	GDP => ELC
Al-Mulali and Che Sab [52]	Sub-Saharan African countries (1971–2009)	EC <=> GDP
Islam et al. [53]	Malaysia (1960–2007)	EC <=> GDP (Short and long run)
Menegaki [45]	Europe (1997–2007)	GDP ≠ RE
Oh and Lee [54]	South Korea (1970–1999)	GDP <=> EC
Fallahi [55]	USA (1960–2005)	GDP <=> EC
Tsani [56]	Greece (1960–2006)	GDP => EC (negatively in high income)
Belke et al. [57]	OECD countries (1981–2007)	EC => GDP
Shahbaz et al. [58]	Portugal (1971–2002)	ELC <=> GDP
Shahbaz et al. [59]	BRICS (1970–2015)	EC <=> GDP
Pirloge and Cicea [60]	Spain, Romania, European Union (1990–2010)	EC => GDP
Armeanu et al. [61]	European Union (2003–2014)	GDP => REC

Notes: => indicates unidirectional relationship, <=> indicates bidirectional relationship, ≠ indicates no causal relationship, FEC—final energy consumption, NEC—nuclear energy consumption, EC—energy consumption, REC—renewable energy consumption, ELC—electricity consumption, GDP—gross domestic product.

provides an overview of studies examining the relationship between energy consumption (both renewable and non-renewable) and economic growth across various countries and regions. Most studies support the “feedback hypothesis”, indicating a bidirectional relationship between energy consumption and GDP, as seen in regions such as Sub-Saharan Africa, OECD countries, and BRICS nations. Some studies, like those in Bangladesh and Greece, suggest the “conservation hypothesis”, where GDP impacts energy consumption. Conversely, the “neutrality hypothesis” was confirmed in Europe, showing no significant link between renewable energy and GDP. Additionally, findings in OECD countries support the “growth hypothesis”, where energy consumption drives economic growth. This inconsistency underscores the need for region-specific analyses.

The varying results across these studies highlight the complexity of the energy–growth relationship. Recent studies offer significant insights into this dynamic, suggesting that while renewable energy consumption generally fosters economic growth, its impact can vary depending on regional and economic contexts. Policymakers need to consider these intricacies when formulating strategies to advance sustainable energy use and economic development. Other factors contributing to these differences include regional economic structures, levels of development, energy policies, and the proportion of renewable energy in the energy mix. Methodological variations, such as data periods and econometric techniques, also influence the outcomes.

Despite extensive research, a paucity of studies focuses specifically on the European Union during its green transition period. The EU’s unique policy environment, characterized by forceful renewable energy targets and climate change initiatives, presents a distinct context for examining the energy–growth nexus. This study aims to fill this gap by providing a long-term perspective on the relationship between renewable and non-renewable energy consumption and economic growth in EU countries from 1995 to 2021.

3. Data and Methods

3.1. Data and Model Specification

This study investigates the relationship between economic growth, gross fixed-capital formation, total labor force, and energy consumption (both renewable and non-renewable) within the European Union. We utilize a balanced panel dataset covering 28 EU countries, including the United Kingdom, over the 27 years from 1995 to 2021. To systematically analyze the data and achieve our research objectives, we follow a structured estimation procedure outlined in

Table 2. Estimation procedure.

Step	Description
Step 1	Data Collection and Preparation -- Variables: GDP, Renewable Energy Consumption, Non-Renewable Energy Consumption, Capital, Labor -- Time Period: 1995–2021 -- Countries: 28 EU countries (including the UK)
Step 2	Preliminary Diagnostic Tests -- Cross-Sectional Dependence Test (Pesaran’s CD Test) -- Slope Homogeneity Test (Pesaran and Yamagata Delta Test)
Step 3	Stationarity Tests (Unit Root Tests) -- Cross-sectional Augmented Dickey–Fuller Test (CADF) -- Cross-sectional Im, Pesaran, and Shin Test (CIPS)
Step 4	Cointegration Tests -- Westerlund cointegration test -- Westerlund error-correction-based panel cointegration tests
Step 5	Estimation of Long-Run Coefficients -- Mean Group (MG) Estimator -- Augmented Mean Group (AMG) Estimator -- Fully Modified OLS (FMOLS) -- Fixed-effect Panel Threshold Model
Step 6	Granger Causality Analysis -- Dumitrescu and Hurlin Causality Test

.

The corresponding variables are gross domestic product measured in a million constant 2015 USD ($GD{P}_{it}$), capturing the economic output, gross fixed-capital formation measured in million constant 2015 USD ($GFC{F}_{it}$), reflecting the investments in fixed assets, total labor force ($L{F}_{it}$), representing the available workforce, renewable energy consumption quantified in billion (one followed by twelve zeros) Btu (British thermal units) ($RE{C}_{it}$), and non-renewable energy consumption measured in billion (one followed by twelve zeros) Btu ($NRE{C}_{it}$), including energy derived from coal, natural gas, petroleum, and other liquids. The data are sourced from the World Bank database.

Figure 1 shows the trends in the mean values of these variables over time (1995–2021). The upward GDP and GFCF trends suggest economic growth and increased investment. Renewable energy consumption is clearly on the rise, while non-renewable energy consumption shows some variability and a decline in recent years, indicating a shift towards more sustainable energy sources.

Table 3. Descriptive statistics.

Variable	Mean	Std. Dev.	Min.	Max.
${\mathrm{GDP}}_{\mathrm{it}}$	540,752.5	839,899.2	4659.17	359,7317
${\mathrm{GFCF}}_{\mathrm{it}}$	109,928	166,826.3	0	752,844.8
${\mathrm{LF}}_{\mathrm{it}}$	8,542,400	1.09 × 107	147,038	4.44 × 107
${\mathrm{REC}}_{\mathrm{it}}$	233.61	347.09	−168.21	2187.1
${\mathrm{NREC}}_{\mathrm{it}}$	2077.49	2824.95	33.58	13,014.55

Note: number of obs. 756.

presents the descriptive statistics for all variables, providing an overview of their distributions across countries and over time. The dataset comprises 756 observations. The variability observed in the data underscores the heterogeneity among EU member states in terms of economic output and energy consumption patterns. The data appear to be right-skewed, influenced by countries with significantly larger economies or energy consumption. The transformation of variables in our case (using logarithms) helps stabilize variance and normalize distributions, which is already considered in the model specification. Due to the presence of negative values in the renewable energy consumption (${\mathrm{REC}}_{\mathrm{it}}$) variable, a special procedure was implemented during the logarithmic transformation to accommodate these values.

To provide insights into how these variables interact within the context of the EU’s evolving green transition, energy policies, and economic frameworks, this study employs the log-linear regression model recently developed by Mahendru, Tiwari, et al. [62].

$$\ln GD{P}_{it}={\beta }_0+{\beta }_1\ln GFC{F}_{it}+{\beta }_2\ln L{F}_{it}+{\beta }_3\ln RE{C}_{it}+{\beta }_4\ln NRE{C}_{it}+{\epsilon }_{it}$$

(1)

$\ln GD{P}_{it}={\beta }_0+{\beta }_1$ where $\mathit{ln}GD{P}_{it}$ represents the natural logarithm of gross domestic products, $\mathit{ln}GFC{F}_{it}$ the natural logarithm of gross fixed-capital formation, $\mathit{ln}L{F}_{it}$ represents the natural logarithm of the labor force, $\mathit{ln}RE{C}_{it}$ is the natural logarithm of renewable energy consumption, and $\mathit{ln}NRE{C}_{it}$ is the natural logarithm of non-renewable energy consumption.

3.2. Methods

We began our panel data analysis by checking for cross-sectional dependence among the variables with Pesaran’s CD test [63], which revealed significant correlations across counties. The growing body of panel data research has focused heavily on the issue of cross-sectional dependence on macro panel data. This correlation may have emerged due to the global financial crisis beginning in 2007 or from economic integration within Europe, possibly driven by local spillover effects. The Pesaran CD test uses the correlation coefficients between the time series of each panel member. For instance, in a dataset with $N=28$ countries, this would involve calculating the $28\times 27$ correlations between country $i$ and all other countries, where $i$ ranges from 1 to $N-1$. Denoting these estimated correlation coefficients between the time series of countries $i$ and $j$ as ${\rho }_{ij}^{*}$, the Pesaran CD statistic is then calculated as follows:

$$\mathrm{CD}=\sqrt{{\displaystyle \frac{2}{\left(\mathrm{N}\left(\mathrm{N}-1\right)\right)}}}\times \left({\displaystyle \sum }_{\mathrm{i}=0}^{\mathrm{N}-1}{\displaystyle \sum }_{\mathrm{j}=\mathrm{i}+1}^{\mathrm{N}}\sqrt{{\mathrm{T}}_{\mathrm{ij}}{\mathsf{\rho }}_{\mathrm{ij}}}\right)$$

(2)

where ${\mathrm{T}}_{\mathrm{ij}}$ represents the number of observations used to compute the correlation coefficient. Under the null hypothesis of cross-sectional independence, the statistics mentioned are normally distributed if ${\mathrm{T}}_{\mathrm{ij}}>3$ and $\mathrm{N}$ is sufficiently large. The test is robust to nonstationarity (since any spurious effects would be averaged out), parameter heterogeneity, or structural breaks, and has been demonstrated to perform well even with small sample sizes.

Currently, the European Union consists of 27 diverse countries (with the United Kingdom also included in our sample), which makes it essential to test the relationship between dependent and independent variables to ensure consistency across all panels in our dataset. This testing is vital in panel data analysis as it helps select the appropriate econometric model. We utilized the Pesaran and Yamagata [64] Delta test ($\Delta$-test) to assess slope homogeneity, which examine whether the slope coefficients are identical across different cross-sectional countries. This test is widely used to evaluate slope homogeneity in panel data models, testing the null hypothesis that the slope coefficients are homogeneous against the alternative hypothesis that they are heterogeneous.

To assess stationarity in our panel data, we employed the Cross-sectional Augmented Dickey–Fuller (CADF) test, proposed by Pesaran [65], and Cross-sectional Im, Pesaran, and Shin (CIPS) test, also introduced by Pesaran [66]. Both tests are econometric methods used to detect unit roots in panel data, especially when cross-sectional dependence is present. The CADF test performs a t-test for unit roots in heterogeneous panels with cross-sectional dependence. To account for cross-dependence, it enhances standard Dickey–Fuller (or Augmented Dickey–Fuller) regressions by including cross-sectional averages of lagged levels and first differences of individual series, resulting in the CADF statistics. In addition, the CIPS test extends the IPS test by incorporating cross-sectional averages to handle cross-sectional dependence better.

Next, we assessed cointegration to determine if the variables share a stable long-term relationship. We employed the Westerlund cointegration test [67], which provides two VR test statistics to test the null hypothesis of no cointegration. We also applied Westerlund’s [68] error-correction-based panel cointegration tests, which include the $Ga$, $Gt$, $Pa$, and $Pt$ tests. The $Ga$ and $Gt$ tests check if at least one panel member shows cointegration ($H_0: {\alpha }_i=0 \mathrm{vs}. H_1: {\alpha }_i<0$ for at least one $i$). The $Pa$ and $Pt$ tests assess cointegration for the entire panel ($H_0: {\alpha }_i=0 \mathrm{vs}. H_1: {\alpha }_i<0$ for all $i$). These tests handle heterogeneous models and varying series lengths, with bootstrap iterations (1000) used for robustness if cross-sectional units are correlated.

In the next step, we applied various econometric techniques, which allow for heterogeneous slope coefficients across group members and are also concerned with correlation across panel members: Mean Group (MG) [69], Augmented Mean Group (AMG) (Bond and Eberhardt [70]); Eberhardt and Teal [71], Common Correlated Effects Mean Group (CCEMG) (Pesaran [72]), Fully Modified Ordinary Least Squares (FMOLS) (Philips and Hansen [73]), Dynamic Ordinary Least Squares (DOLS) (Stock and Watso [74]), and Canonical Cointegrating Regression (CCR) (Park [75]).

Additionally, we estimated a non-linear panel data model and tested for the relationship within that framework. Our fixed-effect panel threshold model (Equations (3) and (4)) is based on the method proposed by Hansen [76] and Wang [77]. The model allows the coefficient of non-renewable energy consumption ($\mathit{ln}NRE{C}_{it}$) to differ depending on whether the threshold variable $\mathit{ln}NRE{C}_{it}$ is below or above the estimated threshold γ. Non-renewable energy consumption might have different impacts on economic growth at different levels due to factors like resource depletion, environmental regulations, or technological changes. The model can be written in two regimes.

Regime 1: when $\mathit{ln}NRE{C}_{it}<\mathsf{\gamma }$

$$\ln GD{P}_{it}={\mu }_i+{\beta }_1\ln RE{C}_{it}+{\beta }_2\ln L{F}_{it}+{\beta }_3\ln GFC{F}_{it}+{\delta }_1\ln NRE{C}_{it}+{\epsilon }_{it}$$

(3)

Regime 2: when $\mathit{ln}NRE{C}_{it}\ge \mathsf{\gamma }$

$$\ln GD{P}_{it}={\mu }_i+{\beta }_1\ln RE{C}_{it}+{\beta }_2\ln L{F}_{it}+{\beta }_3\ln GFC{F}_{it}+{\delta }_2\ln NRE{C}_{it}+{\epsilon }_{it}$$

(4)

where $\ln RE{C}_{it}$, $\ln L{F}_{it}$, and $GFC{F}_{it}$ are regime-independent variables; ${\beta }_1$—effect of renewable energy consumption on GDP, assumed constant across both regimes; ${\beta }_2$—effect of labor force on GDP, assumed constant across both regimes; ${\beta }_3$—effect of gross fixed capital formation on GDP, assumed constant across both regimes; $\mathit{ln}NRE{C}_{it}$—associated with ${\delta }_1$ and ${\delta }_2$ is a regime-dependent variable; ${\delta }_1$—effect of non-renewable energy consumption on GDP when $\mathit{ln}NRE{C}_{it}<\gamma$ (regime 1); and ${\delta }_2$—effect of non-renewable energy consumption on GDP when $\mathit{ln}NRE{C}_{it}\ge \gamma$ (regime 2).

In this model, $\mathit{ln}NRE{C}_{it}$ is a threshold variable, serving dual roles, and determines the regime by comparing $\mathit{ln}NRE{C}_{it}$ to the threshold $\gamma$. ${\mu }_i$ is the country fixed effects which controls for unobserved, time-invariant heterogeneity across countries and captures factors specific to each country that do not change over time (e.g., culture); ${\epsilon }_{it}$ is error term.

In the final stage, we assessed causality, specifically whether one time-series can predict or drive changes in another, using the Granger non-causality test developed by Dumitrescu and Hurling [78]. This approach extends the Granger causality test to panel data, enabling the examination of causality across multiple cross-sectional units (such as countries) while considering potential heterogeneity among these units. The test’s null hypothesis asserts that no causality exists for any cross-sectional units. To account for cross-sectional dependence, we employed a bootstrap procedure to calculate p-values and critical values, as recommended by Dumitrescu and Hurlin [78].

4. Results and Discussion

4.1. Cross-Sectional Dependence and Slope Heterogeneity

Controlling cross-sectional dependence is crucial in a panel data analysis. Cross-sectional dependence can lead to significant issues in choosing econometric models and interpreting results. To detect cross-sectional dependence, we employ Pesaran’s CD test. According to the results shown in

Table 4. Pesaran’s cross-sectional dependence test results.

Variable	CD-Test	p-Value	Corr	Abs (Corr)
$\ln {\mathrm{GDP}}_{\mathrm{it}}$	88.99	0.000	0.881	0.881
$\ln {\mathrm{GFCF}}_{\mathrm{it}}$	63.88	0.000	0.632	0.685
$\ln {\mathrm{LF}}_{\mathrm{it}}$	28.59	0.000	0.283	0.742
$\ln {\mathrm{REC}}_{\mathrm{it}}$	57.79	0.000	0.572	0.576

Note: under the null hypothesis of cross-section independence CD ~ N(0, 1).

, all variables exhibit significant cross-sectional dependence, indicating that the units are not independent.

As we proceed,

Table 5. Results of slope heterogeneity test.

	Statistics	p-Value
Delta	19.527	0.000
Delta Adj.	22.688	0.000

Note: the null hypothesis is slope coefficients are homogenous.

presents the results of the slope homogeneity test, another critical aspect of panel data analysis. According to the results of the Delta test (Δ-test) the slope coefficients are not homogeneous across the units in the panel. This implies that the relationship between the dependent variable ($GD{P}_{it}$) and independent variables ($GFC{F}_{it}$, $L{F}_{it}$, $RE{C}_{it}$) varies across the different countries in the panel.

Considering the presence of cross-sectional dependency and slope heterogeneity indicated by the tests above, we assess the variables’ stationarity, or whether they contain a unit root, using the Cross-sectional Augmented Dickey–Fuller (CADF) and Cross-sectional Im, Pesaran, and Shin (CIPS) unit root tests. From the results in

Table 6. CADF and CIPS unit root tests results.

Variable	CADF		CIPS
	Level	First Diff.	Level	First Diff.
$\ln {\mathrm{GDP}}_{\mathrm{it}}$	−2.106 **	−2.702 ***	−2.020	−3.640 ***
$\ln {\mathrm{GFCF}}_{\mathrm{it}}$	−1.944	−3.728 ***	−1.314	−4.263 ***
$\ln {\mathrm{LF}}_{\mathrm{it}}$	−1.875	−2.795 ***	−1.925	−3.998 ***
$\ln {\mathrm{REC}}_{\mathrm{it}}$	−3.075 ***	−4.514 ***	−2.893 ***	−5.308 ***
$\ln {\mathrm{NREC}}_{\mathrm{it}}$	−2.153 **	−3.819 ***	−2.269 **	−5.160 ***

Note: **, *** denote statistical significance at the 5% and 1% levels, respectively.

, it can be seen that all variables become stationary after first differencing. $\mathit{ln}RE{C}_{it}$ and $\mathit{ln}NRE{C}_{it}$ show some evidence of stationarity even at the level, with stronger evidence after the first differencing.

4.2. Cointegration Tests

We employ the Westerlund cointegration test to examine the long-term relationship between the variables, considering cross-sectional dependency and slope heterogeneity. The results in

Table 7. Westerlund cointegration test results.

	Statistic	p-Value
Variance Ratio	−1.5655	0.0587

Notes: the null hypothesis is that there is no cointegration; cross-sectional means are removed.

indicate a p-value of 0.0587, slightly above the conventional significance level of 0.05. Given its proximity to 0.05, one could argue that there is weak evidence against the null hypothesis. At a 10% significance level, we could reject the null hypothesis and infer the presence of cointegration.

We perform additional tests to enhance the robustness of our findings.

Table 8. Westerlund error-correction-based panel cointegration test results.

Statistic	Value	Z-Value	Robust p-Value
${\mathrm{G}}_{\mathrm{t}}$	−1.045	4.807	0.000
${\mathrm{G}}_{\mathrm{a}}$	−1.334	6.358	0.000
${\mathrm{P}}_{\mathrm{t}}$	−1.650	5.450	0.000
${\mathrm{P}}_{\mathrm{a}}$	−0.424	4.118	0.000

Notes: the number of bootstrap iterations is 1000. The null hypothesis is that there is no cointegration.

presents the results of four-panel cointegration tests proposed by Westerlund [68]. The robust p-values are all reported as 0.000, providing strong evidence against the null hypothesis. This suggests that cointegration is likely present even when accounting for potential cross-sectional dependence or other factors that could impact the test’s validity.

Thus, the Westerlund cointegration test in the variance ratio and error-correction-based forms indicates significant long-run relationships among the variables, reinforcing the interconnected dynamics between energy consumption and economic growth. Our results differ from the findings of Afonso et al. [79] and Papież et al. [80], who found no cointegration between economic growth and renewable and non-renewable energy consumption in 28 and 26 countries (excluding Malta and Cyprus) of the member states of the European Union, respectively. These differences are attributable to several factors, including methodological advancements, broader data coverage, and the inclusion of more recent policy impacts. Both studies may not fully account for the effects of more recent EU energy policies and the acceleration of renewable energy adoption, which became more prominent after 2015, particularly with the launch of the European Green Deal. Our dataset spans from 1995 to 2021 and includes this critical period. It allows us to capture the evolving relationship between energy consumption (especially renewables) and economic growth, reflecting ongoing transitions that may have been less prominent in earlier periods.

4.3. Long-Run Coefficients

We calculated the long-run coefficients using six different econometric estimation techniques, selecting each method according to its specific underlying assumptions.

Table 9. Long-run coefficients.

Variable	MG	AMG	CCEMG	FMOLS	DOLS	CCR
$\ln {\mathrm{GFCF}}_{\mathrm{it}}$	0.364 ***	0.166 ***	0.137 ***	0.37 (117.68) ***	0.35 (64.00) ***	0.37 (105.49) ***
$\ln {\mathrm{LF}}_{\mathrm{it}}$	0.886 ***	0.293 ***	0.241 **	0.85 (64.25) ***	1.08 (23.32) ***	0.88 (47.56) ***
$\ln {\mathrm{REC}}_{\mathrm{it}}$	0.077 ***	0.024 *	0.018	0.07 (40.83) ***	0.16 (36.16) ***	0.07 (32.96) ***
$\ln {\mathrm{NREC}}_{\mathrm{it}}$	0.033	0.192 ***	0.189 ***	0.01 (10.93) ***	0.08 (15.55) ***	0.00 (9.21) ***
$\mathrm{Cons}$	−5.920	4.277 ***	−10.759

Notes: *, **, *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.

presents the results. $\mathit{ln}GFC{F}_{it}$ consistently exhibits a positive and significant impact across all estimation methods, underscoring the role of gross fixed-capital formation as a robust predictor of $\mathit{ln}GD{P}_{it}$. Similarly, $\mathit{ln}L{F}_{it}$ demonstrates a consistently positive and significant association with $\mathit{ln}GD{P}_{it}$, highlighting the importance of labor force size as a key determinant, in line with classical economic theory. The findings for $\mathit{ln}RE{C}_{it}$ generally indicate a positive relationship with $\mathit{ln}GD{P}_{it}$, although the significance and magnitude of the effect vary across different tests. Notably, the AMG and CCEMG tests provide comparatively weaker evidence.

As for $\mathit{ln}NRE{C}_{it}$, the results are mixed. While the MG test suggests that $\mathit{ln}NRE{C}_{it}$ is not significant, other models indicate a positive and significant relationship, though the effect size is generally smaller compared to the other variables.

Moreover, we employ a non-linear panel data model and conduct causality testing within this framework.

Table 10. Results of threshold regression analysis with $\ln NREC$ defining regimes.

Indep Variable $lnGD{P}_{it}$	Coefficient	Std. Err.	t	P > |t|
${\mathrm{lnREC}}_{\mathrm{it}}$	0.041	0.004	11.08	0.000
$ln{\mathrm{LF}}_{\mathrm{it}}$	0.754	0.054	13.83	0.000
${\mathrm{lnGFCF}}_{\mathrm{it}}$	0.287	0.011	25.00	0.000
regime_${\mathrm{lnNREC}}_{\mathrm{it}}$
low	0.054	0.049	1.11	0.269
high	−0.258	0.038	−6.85	0.000
_cons	−0.762	0.796	−0.96	0.339

Notes: threshold estimator (level = 95): 4.2180, R-squared within = 0.6884, R-squared between = 0.8983, R-squared overall = 0.8939.

presents the results of the panel threshold regression model. Based on R-squared, the model has strong explanatory power, particularly in explaining differences between countries and over time. The coefficient associated with renewable energy consumption is highly significant (p-value < 0.001), indicating a strong positive effect on economic growth across all regimes. A 1% increase in renewable energy consumption is associated with an average 0.0408% increase in GDP. The results are consistent with Dissanayake et al. [81] and Magazzino [82], who highlight the negative effects of high non-renewable energy use on the environment and economy. The labor force and gross capital formation coefficients have the same effect.

The model divides the data into two regimes based on the value of $\ln NREC$: regime 1 (low), $\ln NREC$ < 4.2180, and regime 2 (high), $\ln NREC$ ≥ 4.2180. In the low non-renewable energy consumption regime, the coefficient is not significant (p-value = 0.269), indicating that the effect is not statistically different from zero. In the high non-renewable energy consumption regime (regime 2), the coefficient is highly significant (p-value < 0.001), indicating a strong negative relationship between non-renewable energy consumption and economic growth in this regime. Interestingly, under regime 2, increased non-renewable energy consumption is associated with a decrease in GDP. This means that high levels of non-renewable energy use may lead to environmental harm and negatively impact economic activity. Additionally, negative externalities such as health costs and reduced quality of life may outweigh the benefits of energy consumption at high levels. This issue is addressed in the works of Usman et al. [83] and Tudor et al. [84].

4.4. Causality Analysis

To test causality in the panel dataset, we employ the Dumitrescu and Hurlin [78] Granger causality test since it takes into account heterogeneous causality relationships specific to individual units (countries).

Table 11. Dumitrescu and Hurlin Granger causality test results.

	W-Bar	Z-Bar	p-Value	Z-Bar Tilde	p-Value
$GFC{F}_{it}$ = > $GD{P}_{it}$	1.8934	3.3428	0.0008	2.5340	0.0113
$L{F}_{it}$ = > $GD{P}_{it}$	2.0868	4.0664	0.0000	3.1480	0.0016
$RE{C}_{it}$ = > $GD{P}_{it}$	0.5660	−1.6240	0.1044	−1.6804	0.0929
$NRE{C}_{it}$ = > $GD{P}_{it}$	2.1943	4.4686	0.0000	3.4893	0.0005
$GD{P}_{it}$ = > $GFC{F}_{it}$	3.1655	8.1026	0.0000	6.5728	0.0000
$GD{P}_{it}$ = > $L{F}_{it}$	4.5093	13.1305	0.0000	10.8390	0.0000
$GD{P}_{it}$ = > $RE{C}_{it}$	4.7143	13.8975	0.0000	11.4898	0.0000
$GD{P}_{it}$ = > $NRE{C}_{it}$	4.3367	12.4849	0.0000	10.2912	0.0000

Note: the null hypothesis is that there is no Granger causality from one variable to another across all individuals in the panel.

provides the related results. For the relationship between the renewable energy consumption ($RE{C}_{it}$) and gross domestic product ($GD{P}_{it}$), the p-values (0.1044 and 0.0929) exceed the conventional significance level of 0.05. This indicates insufficient evidence to reject the null hypothesis, suggesting that $RE{C}_{it}$ does not Granger-cause $GD{P}_{it}$ in this context.

The analysis reveals bidirectional Granger causality between $GD{P}_{it}$ and both the gross fixed-capital formation $GFC{F}_{it}$ and labor force $L{F}_{it}$, indicating a mutual predictive relationship between these variables. Furthermore, $GD{P}_{it}$ is found to Granger-cause both renewable and non-renewable energy consumption, while non-renewable energy consumption also Granger-causes $GD{P}_{it}$. However, no such Granger-causal relationship is observed between renewable energy consumption and $GD{P}_{it}$, indicating a lack of predictive power in this direction.

However, the literature presents a range of mixed findings. Marques et al. [14] reveal a bidirectional Granger causality between renewable energy consumption and GDP in France. Apergis and Payne [34] focused on 20 OECD countries over a specific period from 1985 to 2005, finding bidirectional causality between renewable energy consumption and economic growth. Furuoka [85] examined the causal links in newly industrialized countries, finding different energy–growth relationships across countries due to varying levels of renewable energy adoption. In countries like South Africa and Mexico, negative renewable energy shocks influenced GDP positively. Farhani and Shahbaz [86], focusing on MENA countries, found a unidirectional causality from renewable energy to economic growth.

These differences can be attributed to variations in geographical focus, time periods, and methodologies. While Apergis and Payne [34] identified bidirectional causality in OECD countries, and Farhani and Shahbaz [86] found unidirectional causality in MENA countries, both regions and periods differ significantly from our study in the European Union, which focuses on more recent years with more robust renewable energy policies. Furthermore, methodological differences may be an additional reason.

Our findings provide valuable insights into the role of renewable energy in economic growth, building upon and contrasting with prior research in the field. For instance, our results partially confirm the conclusions of M.A. Cetin [87], who studied E-7 countries and identified a long-term cointegration between energy consumption and economic growth alongside short-term causality. Utilizing FMOLS and DOLS methods, Cetin demonstrated robust evidence for these relationships. However, our analysis, centered on EU countries during 1995–2021, underscores a more significant role of renewable energy in long-term economic growth. This distinction can be attributed to disparities in technological development, energy policies, and the EU’s more advanced climate initiatives, such as the European Green Deal.

Similarly, studies on BRICS countries [88] have highlighted the positive influence of electricity consumption, the labor force, and investments on economic growth—paralleling findings from our analysis of EU nations. However, a key difference is that renewable energy plays a much more pronounced role in driving long-term economic expansion in the EU. Further support for renewable energy’s economic and environmental benefits emerges from Mahendru et al. [62], who argue that transitioning to renewables advances both goals. Our study corroborates this conclusion, but also reveals that the impact of renewable energy in EU countries varies significantly depending on GDP levels. In contrast to Apergis and Payne [89], who found evidence of bidirectional causality between renewable energy consumption and economic growth in the Eurasian region, our findings did not identify such short-term causal relationships within the EU. Finally, a study on the green transition involving 25 European countries [85] demonstrated a long-term correlation between renewable energy consumption and economic growth, particularly in nations with higher GDP. These findings align closely with our results, further emphasizing the critical role of renewables in fostering sustainable economic development across the EU.

Our research provides detailed evidence that renewable energy, investment, and labor have differential effects on economic growth in EU countries, which can be directly useful for policy analyses in regions such as BRICS, MENA, and newly industrialized economies. In these regions, that support for investment in renewables can yield significant long-term benefits, especially if accompanied by infrastructure development and policies that support the labor market. In MENA countries, where non-renewable energy sources dominate, our analysis highlights the potential impact of renewable energy investments on stabilizing and balancing economic growth, as well as on reducing the vulnerability of economies to energy commodity price fluctuations.

5. Conclusions and Policy Recommendations

This study examines the relationship between economic growth, the gross fixed-capital formation, total labor force, and energy consumption in the European Union, utilizing a panel dataset of 28 countries (including the United Kingdom) from 1995 to 2021. The results from the Westerlund cointegration test, supported by additional robustness checks, suggest a long-term equilibrium relationship between variables, accounting for cross-sectional dependency and slope heterogeneity.

The relationship between renewable energy consumption and GDP is generally positive, indicating that the increased use of renewable energy is associated with economic growth in the long run. However, the significance and strength of this relationship vary across different econometric models. Notably, the AMG and CCEMG tests provide comparatively weaker evidence of a significant impact. Using $\ln NREC$ as the threshold variable in the regression analysis yields better results. The coefficient associated with renewable energy consumption remains highly significant, demonstrating a strong positive effect on economic growth across all regimes. However, the Dumitrescu and Hurlin [76] causality test finds no significant Granger-causal relationship between renewable energy consumption and GDP. This suggests that while renewable energy consumption is positively linked to economic growth, it does not have a strong predictive influence on GDP within the study period.

The results for non-renewable energy consumption are mixed. While the MG model indicates that non-renewable energy consumption is not significant, other models suggest a positive and significant relationship with GDP, though the effect size is generally smaller compared to other variables like the capital formation and labor force. In the estimation using ln NREC as the threshold variable, non-renewable energy consumption exhibits a regime-dependent effect: below the threshold, its effect on GDP is not significant, whereas above the threshold, it negatively impacts GDP. The Granger causality analysis reveals a bidirectional relationship between non-renewable energy consumption and GDP.

Our results have significant policy recommendations, indicating that promoting renewable energy sources does not hinder economic growth. Moreover, such promotion has the potential to contribute substantially to economic growth in the future. Therefore, in addition to other crucial benefits, such as increased energy security, the development of renewable energy sources does not threaten the economy. This is particularly relevant as many EU countries, including Poland, Romania, Hungary, Bulgaria, Slovakia, and Lithuania, still have underdeveloped renewable energy sectors. This study contributes meaningfully to the ongoing debate on renewable energy promotion within the European Union. It clearly signals to policymakers that supporting this development is essential, given the low risk of negative impacts on economic growth and the high likelihood of achieving significant benefits.

The green transition also brings structural changes, including eliminating subsidies for fossil fuels and introducing instruments such as carbon taxes. An important step is also the establishment of regional funds for research and development, which would enable the effective implementation of innovative technological solutions across the European Union. At the same time, efforts to modernize energy infrastructure must be intensified, particularly the upgrading of transmission networks and energy storage systems, to facilitate the integration of renewable energy sources into national power systems. Investments in smart grids will improve the efficiency of energy transmission and ensure the stability of the supply. It is also vital to tailor energy policies to the specific circumstances of individual member states, considering their level of technological development, share of renewable energy sources in the energy mix, and CO₂ emissions. Such flexibility will enable the more effective achievement of energy transition goals, supporting balanced and cohesive development across the European Union.

Supporting efforts to enhance energy efficiency is also critical from an economic standpoint. Investment in energy-efficient technologies reduces reliance on nonrenewable energy sources and lowers operational costs across various sectors. Improved energy efficiency can lead to substantial reductions in greenhouse gas emissions, potentially decreasing future expenditures related to climate regulations and fostering long-term cost savings.

Even though current direct evidence linking renewable energy development to immediate economic growth is weak, investing in renewable energy remains a strategic choice for long-term economic stability. Increased financial support and streamlined regulatory processes are necessary to expedite the transition to modern energy sources, thereby reducing dependence on volatile fossil fuel markets and contributing to more stable and predictable energy costs. At the same time, in EU countries that still rely heavily on fossil fuels, it is essential to raise public awareness about the benefits of energy transition. Enhanced societal awareness can drive legislative and market changes, facilitate the diversification of energy sources, and mitigate risks associated with global price fluctuations in raw materials. In the long run, this approach will strengthen energy security and minimize potential negative economic impacts. Finally, expanding energy integration within the European Union through increased regional cooperation and interconnected energy grids is crucial. Such integration will optimize energy resource utilization, bolster energy security and market stability, and reduce the costs associated with adapting economies to the evolving dynamics of global energy markets.

The findings of this study open several avenues for future research. The results of the slope homogeneity test indicate that the relationship between economic growth and the independent variables varies across different countries within the panel. This heterogeneity implies that aggregating data may mask important country-specific dynamics. Therefore, future research should consider conducting analyses at the cluster level or on a per-country basis to capture these relationships more accurately. To deepen the understanding of these relationships, future studies could employ dynamic panel data models. Additionally, performing robustness checks using alternative control variables would strengthen the validity of the results, ensuring that the findings are not artifacts of specific model specifications.

Analyzing the effects across different regions or income groups can also shed light on whether the observed relationships hold universally or are context-specific. This stratification can help identify whether policies need to be tailored to particular economic environments. The significant fixed effects found in the fixed-effect panel threshold model underscore the importance of unobserved country-specific factors. Recognizing and incorporating these factors into future models will enhance the precision of the analyses and the relevance of the policy implications derived from them.