How does the Energy Transition Shape Inclusive Green Growth in the European Union?

2.1 Energy Transition and Economic Sustainability

The detrimental impact of fossil fuels on the environment urges countries worldwide to shift their energy consumption to renewable energy. This leads the researchers to investigate the linkage between renewable energy transition and economic growth. Apergis and Payne (2010) explored the impact of renewable energy consumption on economic growth in 20 Organisation for Economic Co-operation and Development (OECD) countries from 1985 to 2005. The study confirmed the positive impact of renewable energy on economic growth for those countries. Another study by Apergis and Payne (2011) established the profound impact of renewable energy usage on economic growth for six Central American countries. For 34 OECD countries, Inglesi-Lotz (2016) affirmed that total renewable energy consumption and an increase in the share of renewable energy consumption both have a promotional effect on economic growth.

Bhattacharya et al. (2016) studied the impact of renewable energy consumption on economic growth in the top 38 renewable energy-consuming countries spanning 1991–2012. The findings showed that renewable energy enhanced the prosperity of those economies. Similarly, Rafindadi and Ozturk (2017) asserted that renewable energy consumption increased Germany’s per capita GDP. Further, in a study of 103 countries, Chen et al (2020) found that renewable energy consumption positively affected economic growth in OECD countries. However, renewable energy amplified economic growth in non-OECD and developing countries after a certain threshold period. In addition, Jan et al (2021) noticed that renewable energy spurred economic prosperity more efficiently than other sources in Pakistan. Similar kind of findings can be observed from the studies by Iqbal et al (2023) and Z. Wang et al. (2021) for BRICS countries and 10 Asian countries, respectively.

However, some studies have different opinions. In a Turkish survey, Ocal and Aslan (2013) found that an increase in the share of renewable energy consumption in the total final energy consumption hampered economic growth. These findings are supported by the studies of Maji et al (2019) and Tenaw (2022) in the case of 15 West African countries and 20 Sub-Saharan countries, respectively. At the initial stage, the development of renewable energy technologies is associated with higher costs, leading to higher energy prices. This discourages people from adopting it and thus negatively affects the economy. Therefore, to get the beneficial impacts of renewable energy, the deployment of renewable energy must cross a critical threshold level.

2.2 Energy Transition and Environmental Sustainability

Studies across different regions and periods unanimously argued that using renewable energy brings environmental sustainability. During the analysis, they used different indicators of environmental degradation, such as CO₂ emissions, ecological footprint and carbon footprint. Gill et al (2018) evaluated the influence of an increase in the share of renewable energy production on CO₂ emissions in Malaysia over the period 1970–2011. The study revealed that renewable energy reduced CO₂ emissions in Malaysia. Similar findings can be observed in a study by Murshed et al. (2021), which indicated that non-fossil fuel and hydroelectricity consumption curbed the carbon footprint in Bangladesh. Further, in a study of 15 Asian economies, Anwar et al. (2022) explained that an increase in the share of renewable energy consumption enhanced environmental quality. Afshan et al (2022) constructed a composite energy transition index and scrutinised the impact of the energy transition on the ecological footprint in 27 OECD countries during 1990–2014. The study confirmed that energy transition promoted environmental sustainability by curbing ecological footprint.

In addition, Kazemzadeh et al. (2024) explored the impact of energy transition on the environment using the ratio of renewable energy to non-renewable energy consumption and observed that energy transition reduced environmental degradation. The study further stated that the impact of energy transition is more intense on higher quantiles of CO₂ emissions. The same findings can be noticed in a study by Liao et al. (2023) for 10 OECD countries. Moreover, studies by Gao and Chen (2023) and Salahodjaev et al. (2022) reported that an increase in the share of renewable electricity production augmented environmental quality in 45 Europe and Central Asia countries and 21 industrialised countries, respectively. A study by Sadiq et al. (2023) established that renewable energy mitigated environmental degradation in BRIC countries. Other studies shared the same vision (Ahmad et al. 2023; Gu and Liu 2023; Kongkuah 2024; Yang et al. 2023). Therefore, from the above discussion of literature, it can be confirmed that the transition from non-renewable to renewable energy makes the environment sustainable.

2.3 Energy Transition and Social Sustainability

While plenty of researchers focused on finding the impact of renewable energy transition on the economy and environment, few studies delved into the social impact of renewable energy transition. Apergis and Salim (2015) studied the impact of renewable energy consumption on unemployment in a panel of 80 countries from 1990 to 2013. The study reported that renewable energy consumption increased unemployment for the overall panel, though the results varied across different regions. However, these findings are inconsistent with the findings of the studies by Khobai et al. (2020) and Naqvi et al (2022) for South Africa and 10 European countries, respectively. The studies found that renewable energy usage reduced unemployment in these economies. In addition, Ram et al (2020) performed an analytical job creation assessment for the global power sector from 2015 to 2050 and reported that 100 % electricity generation from renewable sources by 2050 would create job opportunities from about 21 million in 2015 to nearly 35 million in 2050.

Further, Topcu and Can (2020), in a study of 23 developed economies, revealed that increasing the share of renewable energy consumption reduced income inequality. Sasmaz et al. (2020) explored the impact of renewable energy on the human development index in 28 OECD countries over the period 1990–2017. The study reported an increase in the share of renewable energy improved human development in these economies. The findings are similar to the findings of the studies by Kaewnern et al. (2023) and Z. Wang et al. (2021) for BRICS countries and the top 10 human development countries, respectively. In a study for India, Mamidi et al. (2021) found that the transition to clean energy amplified household development. On the other hand, Iddrisu et al (2023) and Nketia et al. (2022) indicated that increasing the share of renewable energy consumption had a negative impact on inclusive growth across 48 African and 30 Sub-Saharan African (SSA) countries, respectively. Likewise, in the case of economic sustainability, the deployment of renewable energy must reach beyond a certain threshold level to attain social sustainability.

2.4 Other Factors Affecting IGG

Along with renewable energy transition, other variables can influence IGG directly or indirectly. Ghourchian and Yilmazkuday (2020) carried out a study on the impact of government consumption expenditure on economic growth in a group of 83 countries during 1960–2014. The findings of the study established that government consumption expenditure hampered economic growth. Further, a study by Onifade et al. (2020) also confirmed that government recurrent expenditures negatively affected economic growth in Nigeria from 1981 to 2017. The effect of government expenditure on income disparity was examined by Sidek (2021) over a sample of 122 countries and revealed that government expenditure reduced the income gap in those countries. Pham (2024), in his research of 35 Asian countries spanning from 2000 to 2022, documented that government expenditure diminished the unemployment rate. Further, Le and Ozturk (2020) showed that government expenditure degraded the environmental quality by enhancing economic activities and luring more trade and investment activities in 47 emerging markets and developing economies. However, a study by Bilal et al. (2022) confirmed that government expenditures and inflation improved environmental quality in Germany spanning 1971–2016. Baharumshah et al (2016) showed that inflation dampened economic growth in 94 developing and emerging countries from 1976 to 2010.

Conversely, for 79 developing countries, Uddin and Rahman (2023) reported that inflation boosted spending and investment, further stimulating economic growth. In addition, Law and Soon (2020) examined the association between inflation and income inequality for 65 developed and developing economies over the period 1987–2014. The study revealed that inflation exacerbated income inequality in those countries. A study by Rahman et al. (2022) showed that inflation increased CO₂ emissions in Pakistan. Furthermore, Ofori and Figari (2023) explored the impact of economic globalisation on IGG in 23 African countries from 2000 to 2020. They found that economic globalisation negatively affected IGG in those countries. Cioacă et al. (2020) showed that information and communication technology promoted sustainable development in 28 EU countries by stimulating economic growth and reducing income inequality. A study by Ofori et al (2022) revealed that information and communication technologies (ICT) fostered IGG while financial deepening hampered it in 23 African countries from 2000 to 2020. The study further revealed that trade openness deteriorated environmental quality in those economies. Similarly, a study by Xin et al. (2023) concluded that the digital economy promoted IGG in China. However, another study by Ofori et al (2023) reported that ICT harmed IGG, whereas financial development promoted it in 20 SSA countries over the period 2000–2020. Henceforth, it can be concluded that the effect of ICT and financial development depends on the policy framework of countries. The study further concluded that foreign direct investment (FDI) and trade openness enhanced environmental degradation.

2.5 Literature Gap

The review of related literature reveals that numerous studies have examined the influence of energy transition on the three dimensions of sustainability – social, economic and environmental – individually. However, there is a notable lack of research addressing the impact of energy transition on IGG, which integrates all three dimensions to provide a comprehensive view of sustainability. This gap leaves a void in understanding the overall impact of energy transition on sustainable development.

Additionally, while several studies have explored the factors influencing IGG, particularly in regions such as China and Africa, similar research is conspicuously absent for EU nations despite their significant role in advancing global sustainability. Moreover, prior studies often rely on a single indicator to represent energy transition, failing to capture its dynamic and multifaceted nature.

In order to address these gaps, this study constructs composite indices for both IGG and renewable energy transition and investigates the impact of renewable energy transition on IGG across 25 EU nations over the period 1995–2021. This approach provides a more nuanced understanding of the relationship between energy transition and sustainable development within the EU context.

4.1 Data Sources and Variable Construction

The study employs annual data from 25 EU nations^[1] over the period 1995–2021. The selection of the time period and countries is based on data availability. The outcome variable of our study is IGG. The data of IGG are not directly available. We have formulated a composite index of IGG by following the study of Ofori et al (2022). This index incorporates sustainable development’s social, economic and environmental perspectives. The detailed method of constructing the IGG index is described in the next section. The key explanatory variable of our study is the renewable energy transition (RET). Previous studies have taken different proxies of energy transition – the ratio between electricity generated by renewable energy and fossil fuels (Afonso et al 2021), renewable energy consumption and generation (Shahbaz et al. 2022) and the share of primary energy from renewable energy sources (Dogan et al. 2022).

However, according to a report by the International Energy Agency (IEA 2019), one indicator cannot be enough to grasp the complexity of energy transition to clean energy. Therefore, we have created a composite index of RET by incorporating the indicators suggested by the IEA. Apart from energy transition, there are other factors that can influence IGG, and if we do not control their effects, this may lead to biased results. Therefore, the study includes financial development (FD), government expenditure (GE), inflation rate (IR), economic globalisation (EG) and internet access (INTR) as control variables. They are opted from the previous literature on IGG (Ghourchian and Yilmazkuday 2020; Le and Ozturk 2020; Law and Soon 2020; Ofori et al 2022; Ofori and Figari 2023; Xin et al. 2023). The description and data sources of the variables (including energy transition indicators) are provided in Table 2. All the variables are transformed into natural logarithms. This will reduce the sharpness of the data and express the coefficients in terms of elasticity.

4.2 Construction of IGG Index

The term IGG was first introduced during the 2012 G20 Summit under Mexico’s presidency. IGG represents a development paradigm that harmonises economic growth, social equity and environmental sustainability (GGKP 2016; Jha et al 2018). In simpler terms, IGG embodies the integration of these three pillars to foster a balanced and inclusive approach to growth (Ofori et al 2022; Wu et al. 2024). According to the Global Commission on the Economy and Climate, achieving ‘inclusive, high-quality, and resilient’ growth is a critical priority for sustainable development (Morgan 2014). This notion of ‘better growth’ encompasses raising incomes, alleviating poverty, improving public health, making cities more livable, enhancing resilience, promoting innovation and reducing GHG emissions.

Given its centrality to sustainable development, understanding the factors that drive IGG across economic, environmental and social dimensions is crucial. The United Nations’ adoption of the 17 sustainable development goals (SDGs) in 2015 underscores the urgency of achieving sustainable growth by 2030 (UN 2015). However, quantifying IGG poses challenges due to its complexity and multidimensional nature. Researchers have primarily adopted two broad methodological approaches to construct Inclusive Green Growth (IGG) indices: Data Envelopment Analysis (DEA) and index-based aggregation methods. Each of these approaches offers distinct conceptual and empirical advantages depending on the objective and scope of the study. The DEA approach relies on a non-parametric input–output framework using linear programming to estimate efficiency frontiers. One of its key strengths lies in its ability to accommodate multiple inputs and outputs without assuming a predefined functional form. In the context of IGG, this method has been advanced through the use of the Generalised Malmquist–Luenberger index, which incorporates both desirable outputs (such as economic growth) and undesirable outputs (such as environmental degradation and social costs) (Oh 2010; Ren et al. 2022; Zhu and Ye 2018). This framework enables a dynamic assessment of green growth by capturing both improvements in efficiency and shifts in the technology frontier over time.

In contrast, index-based approaches – as exemplified by Jha et al (2018) in their IGG index developed for the Asian Development Bank – rely on aggregating a set of pre-selected indicators, often aligned with the SDGs. These indicators typically span economic, social and environmental dimensions and are combined using either equal or weighted scoring techniques. While such indices offer valuable cross-country comparability and policy communication advantages, they are often sensitive to variable selection and weighting schemes. This study employs the indexing method, leveraging Principal Component Analysis (PCA). PCA is a robust dimensionality reduction technique that addresses collinearity among the 23 selected indicators while retaining their multi-dimensionality. This method produces a set of principal components (PCs), enabling the derivation of a unified IGG index from diverse indicators. Before constructing the index, the selected indicators, grouped into social, economic and environmental categories, are discussed below.

Social indicators – evaluate a nation’s inclusiveness. Access to improved sanitation and potable water signifies a population’s ability to meet basic needs. Higher access rates reflect greater inclusivity. Conversely, increased population density often exacerbates resource exploitation, environmental degradation, pollution and disease transmission, ultimately lowering the quality of life. Thus, population density has a negative correlation with sustainable development. Life expectancy at birth and infant mortality rates are proxies for healthcare quality and well-being. Higher life expectancy and lower infant mortality positively contribute to IGG (Jha et al 2018). Transport infrastructure is also integral to sustainable development. This study uses two indicators as proxies:^[2] air transport, measured by registered carrier departures worldwide, and railway infrastructure, measured by total rail length. Both positively influence the IGG index (Ofori and Figari 2023).

Economic indicators – assess a country’s prosperity. Rising per capita GDP is essential for fostering economic opportunities. However, equitable income distribution is equally important for quality growth. Income inequality, measured by the Gini coefficient, negatively impacts sustainable development, as a higher Gini coefficient reflects income concentration among the wealthy. Human capital, indicative of education quality, is vital for sustainable development. High unemployment rates, however, inhibit IGG by straining economies (Ofori et al. 2022).

Environmental indicators – focus on conservation efforts. While agriculture ensures food security and reduces resource depletion, it can also harm the environment via GHG emissions. This study includes agricultural methane emissions to account for agriculture’s environmental impact. Forests play a pivotal role in environmental conservation by producing oxygen, absorbing CO₂ and preserving biodiversity. Rising global temperatures and higher natural resource rent values, which signify faster depletion rates, negatively impact IGG (Ofori et al 2023). Air pollution’s adverse effects are captured through indicators like mean population exposure to PM2.5, mortality from ambient PM2.5 exposure and welfare costs of premature mortalities (in GDP terms). Carbon productivity, reflecting the decoupling of GDP growth from CO2 emissions, positively contributes to sustainable development. Renewable energy consumption promotes environmental sustainability by replacing fossil fuels, which contribute to global warming and climate change. Environment-friendly technologies further mitigate the negative impacts of human activities and conserve the environment (Li et al. 2021).

To construct the IGG index, we apply Principal Component Analysis (PCA), following the methodology of Kumar et al (2007). This approach allows us to derive weighted indices for each sustainability pillar – social, economic and environmental – by transforming correlated indicators into orthogonal principal components. The data sources and definitions for all indicators used are listed in Table 3, and Table 4 maps these indicators to relevant UN SDGs. Specifically, the social sustainability index comprises seven indicators, the economic index four indicators and the environmental index 12 indicators. These sub-indices are then integrated to construct the overall IGG index.

Before applying PCA, all indicators are normalised using the min-max scaling method to ensure comparability across different units and scales. For each country, PCA is conducted separately to reflect country-specific variation in indicator dynamics, resulting in distinct sets of weights for each national IGG index. This country-specific application of PCA ensures that the resulting indices are sensitive to contextual differences in sustainability patterns. We calculate the index values using the weighted average of all PCs to retain the complete information embedded in the dataset, thereby capturing 100 % of the total variance (for technical details, see Appendix Note A.2). This avoids arbitrary selection of principal components and eliminates the risk of information loss. A flowchart explaining the mechanism of PCA is given in Figure 3.

Figure 3:

Flowchart of obtaining IGG score. Source: Authors’ construct.

The scree plots generated from the PCA applied to Denmark are presented in Figure 4, while the corresponding scatter plots are illustrated in Figure 5. The scree plots display the eigenvalues associated with each principal component, enabling an assessment of their relative importance. The scatter plots, on the other hand, visualise the orthogonality of the principal components. As shown in Figure 5, the principal components are clearly uncorrelated, reaffirming the PCA’s dimensional independence.

Figure 4:

Scree plot of components’ eigenvalue. Source: Authors’ construct.

Figure 5:

Scatterplot of PCs of social, economic and environmental sustainability and IGG. Source: Authors’ construct.

To further aid interpretation, Figure 6 presents biplots for the social, economic, environmental and overall IGG indicators for Denmark, respectively. Each biplot integrates both the scores of individual observations and the loadings of variables onto a single two-dimensional plot. The points in the biplot represent individual observations based on their scores on the first two PCs – where proximity among points indicates similarity in their sustainability profiles. The arrows (vectors) denote the original variables and illustrate both their direction and strength of contribution. Vectors pointing in the same direction suggest positive correlations, while those in opposing directions indicate negative relationships. Longer vectors imply a greater contribution to the explained variance, highlighting variables that exert a stronger influence in shaping each index. Moreover, PCA results for Denmark are detailed in Appendix Tables B.2 –Appendix B.6.^[3] A comparative analysis of IGG scores for 1995 and 2021 is illustrated in Figure 7, showing Czechia’s leading IGG score in 1995 and Denmark’s top position in 2021. Additionally, social, economic and environmental trends for 1995 and 2021 are also visualised in Figure 7.

Figure 6:

Biplot of social, economic, environmental and IGG indicators. Source: Authors’ construct.

Figure 7:

Countrywide IGG scores and its components: 1995 vs. 2021. Source: Authors’ construct.

4.3 Empirical Modelling

The relationship between IGG and energy transition in the presence of control variables can be expressed by the following expression:

where all the variables are previously defined, and subscript i and t are the number of cross sections and time periods, respectively. The logarithmic form of the relationship can be expressed in the following way:

where β ₀ is the intercept, β _i (i = 1,2,…,6) represents the elasticity parameter to be estimated and u _it is the independent and identically distributed error term.

4.4 Econometrics Methods

Before proceeding to estimation, we need to check cross-sectional dependence (CD), which is crucial in studying panel data. Increasing globalisation and trade liberalisation have increased interdependence among countries, and ignorance of this may lead to misleading conclusions (Salahuddin et al 2020). In this study, four CD tests – Breusch–Pagan Lagrange multiplier (LM) test, Pesaran scaled LM test, Pesaran CD test, Bias-adjusted LM test (Breusch and Pagan 1980; Pesaran 2021; Pesaran et al 2008) – are carried out. The results of these tests indicate the presence of CD in the data.

In order to check the time series properties of the variables, the cross-sectionally augmented Im–Pesaran–Shin (CIPS) unit root test (Pesaran 2007) is employed, which considers the CD problem. Before estimating the model, it is required to confirm the presence of a long-run relationship among the variables. Otherwise, it may be a spurious estimation. In order to check cointegration among the concerned variables, the study applies the Pedroni (Pedroni 2004) and the Westerlund (2005) cointegration tests. Further, the study conducts diagnostic tests, including the variance inflation factor (VIF) test for multicollinearity, the Wooldridge test for autocorrelation, the modified Wald test for group-wise heteroskedasticity, the slope homogeneity test developed by Bolmquist and Westerlund^[4] (2013) and the Hausman (1978) test between fixed and random effect model.

The study estimates the empirical relationship among underlying variables using the pooled mean group-autoregressive distributed lag (PMG-ARDL) model proposed by Pesaran et al (1999). The PMG-ARDL model is a dynamic model that can capture the long-run and short-run effects of explanatory variables simultaneously. It is suitable when heterogeneity exists among the panel and allows for the specific lag structure of the variables. One of many appealing features of the model is that it can be applied irrespective of the order of integration of the variables, i.e. when some variables are I(0) and others are I(1) [but not I(2)]. Further, the PMG-ARDL model pools the long-run coefficients while allowing short-run coefficients, intercept and speed of adjustment to vary across panels. Being a dynamic model, it also takes into account the potential endogeneity among the variables.

Following Pesaran et al. (1999), the PMG-ARDL model for this study can be expressed as follows:

Equation (3) can be expressed in the error correction form as follows:

Where ECT _it–1 = LnIGG _it–1− θ ₁LnRET _it–1− θ ₂LnFD _it–1− θ ₃LnGE _it–1− θ ₄LnIR _it–1− θ ₅LnEG _it–1 − θ ₆LnINTR _it–1 and ρ _i is the coefficient of error correction term that indicates the speed of converging the model to long-run equilibrium after a shock in explanatory variables in the short run. Statistical significance of ρ _i ensures the existence of non-linear cointegration among the underlying variables. Moreover, θ 1 = − β 1 ρ , θ 2 = − β 2 ρ , θ 3 = − β 3 ρ , θ 4 = − β 4 ρ , θ 5 = − β 5 ρ , and  θ 6 = − β 6 ρ are the long-run coefficients of LnRET, LnFD, LnGE, LnINF, LnEG and LnINTR, respectively, which are the same across panels and ω _1ij , ω _2ij , ω _3ij , ω _4ij , ω _5ij , and ω _6ij are their short-run coefficients, which vary across panels. However, if the pooling assumption does not hold, the MG estimator is applicable. The MG estimator developed by Pesaran and Smith (1995) allows both the long-run and short-run coefficients to vary across panels. To choose between PMG and MG estimators, we run the Hausman test, where the null hypothesis is the PMG estimator.^[5]

One of the main limitations of PMG-ARDL is that it does not consider CD, which is found in this study. Further, the data suffer from autocorrelation and heteroscedasticity problems. Therefore, in order to check the robustness of the result obtained from PMG-ARDL, the study also employs the fixed effect model with the Driscoll–Kraay standard errors (DKSE) method introduced by Driscoll and Kraay (1998), the feasible generalised least square (FGLS) method advocated by Parks (1967) and the panel-corrected standard errors (PCSE) method pioneered by Beck and Katz (1995). These methods estimate standard errors, which are robust to the presence of CD, autocorrelation and heteroskedasticity problems Hoechle (2007) and thus provide unbiased and consistent outputs. Last, the study carries out the Dumitrescu–Hurlin (D–H) causality test proposed by Dumitrescu and Hurlin (2012) to check the direction of causality among variables. The D–H causality test considers the CD problem. The D–H causality model is illustrated below:

where τ _i denotes the constant value, δ i k represents the autoregressive parameters and λ i k refers to the regression coefficients. The null and alternative hypotheses of the D–H causality test can be defined as follows:

The null hypothesis indicates that there exists no causality for any cross-sectional units in the panel, whereas the alternative hypothesis implies that there exists causality for at least one cross-sectional unit in the panel. A flowchart organising and illustrating the steps involved in the overall framework is presented in Figure 8.

Figure 8:

Flowchart of steps involved. Source: Authors’ construct; Note: IGG represents inclusive green growth; RET represents renewable energy transition; FD represents financial development; GE expenditure represents government expenditure; IR represents inflation rate; EG represents economic globalisation; INTR represents internet access; VIF represents variance inflation factors; CD represents cross-sectional dependence; LM represents Lagrange multiplier; CIPS represents cross-sectionally augmented Im–Pesaran–Shin; PMG-ARDL represents pooled mean group-autoregressive distributed lag; DKSE represents Driscoll–Kraay standard errors; FGLS represents feasible generalised least square; PCSE represents panel-corrected standard errors; D–H represents Dumitrescu–Hurlin.

5.1 Descriptive Statistics

The discussion starts with descriptive statistics of the concerned variables, which are given in Table 5. It is found from the descriptive statistics that IR has the highest variation (SD = 41.683), whereas FD has the lowest (SD = 0.208). This implies that the inflation level largely varies across these countries while they are experiencing similar financial development. Further, IGG, RET, GE and IR are positively skewed, whereas FD, EG and INTR are negatively skewed. The Jarque–Bera statistic shows that all the variables do not follow a normal distribution except GE. From the heat plot of the correlation matrix (Figure 9), it is evident that RET and INTR are positively correlated with IGG, whereas GE and IR are negatively correlated with it. However, FD and EG have no significant correlation with IGG. Further, the correlations between explanatory variables are not high. Therefore, the problem of severe multicollinearity can be precluded from our study, which is more evident from the VIF test (Table 6).

Figure 9:

Heat plot of the correlation matrix. Source: Authors’ construct.

5.2 CD and CIPS Unit Root Tests

After confirming the absence of multicollinearity problems and before proceeding with estimation, we need to check the CD and stationary properties of the variables while dealing with panel data. Tables 7 and 8 display the results of four CD tests and the CIPS unit root test, respectively. The CD test results provide evidence of the presence of CD in data for all variables. Therefore, second-generation methods are applicable to this study. Accordingly, the result of the CIPS unit root test indicates that IGG, RET, FD, IR and INTR are stationary at level, i.e. I(0), whereas GE and EG are stationary at the first difference, i.e. I(1). Since the variables are in mixed order, i.e. I(0) and I(1), and none of the variables are I(2), the PMG-ARDL model is appropriate.

Since all the variables are not stationary at the level, we need to check the existence of a long-run relationship among the underlying variables. For this, the Pedroni (2004) and the Westerlund (2005) cointegration tests were carried out. The results of the panel cointegration tests reported in Table 9 provide evidence for the existence of a long-run relationship among variables.

5.3 PMG-ARDL Estimation

After finding a long-run relationship among the variables, we can now estimate the relationship. But before, pre-diagnostic checks are required. A perusal of the results of pre-diagnostic tests in Table 10 shows that data are not poolable and have autocorrelation and heteroskedastic problems. The result of the slope homogeneity test confirms the presence of heterogeneity among panels. In addition, the result of the Hausman test shows the presence of a fixed effect in the model. Table 11 reports the result of the PMG-ARDL model.

A perusal of Table 11 reveals that the coefficient of the error correction term is significant at the 5 % level, and its value is negative. Therefore, it again confirms the existence of cointegration among the concerned variables. It further implies that after any shock in the explanatory variables, the system converges to equilibrium by 0.038 % annually. In addition, according to the Hausman test, the null hypothesis cannot be rejected, which makes PMG estimators applicable to our study. RET has a negative short-run impact but a positive long-run impact on IGG in the selected EU economies. A 1 % increase in RET results in a 0.038 % decrease in IGG in the short run, whereas a 1.354 % increase in IGG in the long run. The development of renewable energy technologies and plants is initially associated with higher costs and uncertainty, which makes this project less profitable. Therefore, profit-seeking investors remain reluctant to make their investments. Further, because of the expense of renewable energy sources, only richer people can afford them and improve their well-being, which further amplifies the inequality in society. Hence, when countries are transitioning from non-renewable to renewable energy, it can hamper their prosperity and increase inequality, thus ultimately inhibiting IGG (Maji et al. 2019; Tenaw 2022).

However, in the long run, with proper government policies, when the development of the renewable energy industry reaches that threshold level from where large-scale effect starts to operate, costs go down and profits increase. This lures more private investors in this industry and makes renewable energy more available, which further stimulates the economic growth of a country (J. Wang et al. 2023). With the availability of renewable energy sources, prices decrease, and more people can benefit from them. In addition, the consumption of renewable energy reduces GHG emissions, lessens catastrophic climate vents and thus makes the environment more sustainable (Sadiq et al. 2023), which ultimately stimulates the well-being of the people. Further, energy transition creates more job opportunities, reduces income inequality and increases human development (Kaewnern et al. 2023; Ram et al 2020; Topcu and Can 2020). Henceforth, RET has a favourable impact on IGG in the long run.

Considering the control variables, GE hampers IGG in both the long and short run in these selected nations. A 1 % increase in GE reduces IGG by 0.295 % and 0.017 % in the long and short run, respectively. Government spending can retard economic growth if it is more recurrent spending (Onifade et al. 2020). Further, government expenditure degrades environmental quality by raising economic activities (Le and Ozturk 2020). Though public spending on social protection by EU member states is substantially high (Eurostat 2024), which increases inclusiveness in the economy, it may fail to serve the ‘growth’ and ‘green’ part, and the ultimate impact on IGG is detrimental.

Inflation promotes IGG in the short run but deteriorates it in the long run. A 1 % increase in IR increases IGG by 0.002 % in the short run but decreases it by 0.020 % in the long run. Mild inflation is good for the economy as it stimulates investment by creating profit-making opportunities for investors (Uddin and Rahman 2023). However, sustained high inflation hampers economic growth as it reduces purchasing power and creates a dubious environment for investment (Baharumshah et al 2016). Moreover, inflation widens the income gap (Law and Soon 2020) and degrades environmental quality (Rahman et al. 2022). As a result, inflation has a negative impact on IGG in the long run. On the other hand, FD, EG and INTR only have significant long-run effects on IGG in these countries. A 1 % increase in FD and IR inflates IGG by 0.117 % and 0.008 %, respectively, while a 1 % increase in EG reduces IGG by 0.351 %. A developed financial system channels more investments in green projects and brings socio-economic sustainability by including more people into the banking system. Further, access to information can empower individuals, which may result in greater work prospects and higher earnings. This will ultimately boost IGG and aid in reducing social inequality. On the other hand, economic integration of the EU economies with the rest of the world may fail to serve the lower stratum of society or degrade environmental quality, thus lowering IGG. These findings are akin to studies by Ofori and Figari (2023) and Xin et al. (2023).

5.4 DKSE, FGLS and PCSE Estimations

In order to validate the findings of the PMG-ARDL model, the study employs DKSE, FGLS and PCSE methods. These methods are consistent in the presence of autocorrelation, heteroscedasticity and CD. Therefore, they will give robustness to the findings of the PMG-ARDL model. The results are displayed in Table 12. The impact of RET on IGG is found to be positive through all three models, which establishes the consistency of the findings of the study. This statement is also true for the control variables. According to the findings of these three methods, FD and INTR enhance IGG, while GE, IR and EG hinder it. However, all the coefficients are statistically significant only in the FGLS method.

For further sensitivity analysis, we employ the augmented mean group (AMG) estimator introduced by Eberhardt and Teal (2010) and the dynamic common-correlated effects (DCCE) estimator advocated by Chudik and Hashem Pesaran (2015). Both estimators address the problems of nonstationarity, cross-sectional dependence and slope heterogeneity. The results are consistent with those of PMG-ARDL, DKSE, FGLS and PCSE, which strengthens the reliability of our study findings. For further details, see Appendix Table B.8.

5.5 D–H Causality Test

Model estimation does not provide any information regarding the direction of causality among the variables. Therefore, the study further carries out the D–H causality test. Table 13 exhibits the findings of the D–H causality test. The table shows that a bidirectional causality exists between RET and IGG, FD and IGG and EG and IGG. On the other hand, a unidirectional causality runs from IR and INTR to IGG. Further, a unidirectional causality running from IGG to GE has been found.