The Relationship Between Financial Development, Energy Consumption, Economic Growth, and Environmental Degradation: A Comparison of G7 and E7 Countries

Abstract

Both developed and developing countries increased their energy consumption while continuing to advance economically and financially. In parallel with increasing energy use, the intensification of anthropogenic activities has led to higher greenhouse gas emissions, exposing countries to the challenges of climate change and global warming. The environmental degradation resulting from rapid growth in both developed and emerging economies has drawn the interest of scholars, policymakers, and environmental advocates. This study aims to address the relationships between financial development, economic growth, energy consumption, and environmental degradation in G7 and E7 countries. Within this framework, panel cointegration and causality analyses were conducted using annual data from the period between 2000 and 2021 for the relevant countries. The results of the cointegration analysis indicate that the variables move together in the long run in both groups of countries. Furthermore, the long-term relationship coefficients reveal that economic growth and energy consumption contribute to environmental degradation in both G7 and E7 nations. Moreover, the results show that, unlike in E7 countries, financial development in G7 countries exacerbates environmental degradation, while trade openness mitigates it. Panel causality analysis reveals that in E7 countries, changes in financial development influence CO₂ emissions, and variations in CO₂ emissions, in turn, affect economic growth and trade openness. In G7 countries, the analysis results indicate a bidirectional causal relationship between trade openness and CO₂ emissions across the panel. The panel cointegration and causality analyses yield differing results at the country level. Given these findings, it can be recommended that both G7 and E7 countries transition from fossil fuel sources to clean energy sources in conducting economic activities, promote green economy initiatives, and expand the use of green finance instruments to mitigate environmental degradation.

1. Introduction

The Industrial Revolution, rapid population growth, unplanned urbanization, and deforestation, together with the increase in anthropogenic activities, and consequently elevated atmospheric concentrations of greenhouse gases such as carbon dioxide (CO₂), methane (CH₄), sulfur hexafluoride (SF₆), chlorofluorocarbons (CFCs), nitrous oxide (N₂O), and ozone (O₃) triggered global warming (Demir, 2009). The resulting rise in global temperatures has accelerated environmental degradation through phenomena such as forest fires, excessive rainfall, flooding, and erosion (Ureigho, 2018; IPCC, 2014; Vo, 2022). Today, environmental degradation poses a significant threat to global welfare and sustainability. In response, governments have undertaken substantial efforts to mitigate environmental degradation by implementing strict environmental regulations, implementing high environmental taxes, offering financial subsidies and advantageous pricing for the production and consumption of renewable energy, promoting research into energy-efficient technologies, and launching public awareness campaigns to increase environmental consciousness. At the international level, governments have collaborated on the implementation of various environmental agreements (including the 1989 Montreal Protocol on the phasing out of ozone-depleting substances, the 1998 Kyoto Protocol on greenhouse gas emissions reduction, and the 2016 Paris Agreement aiming to limit the increase in global average temperatures) to prevent further deterioration of environmental quality. Despite these efforts, the diverse nature of threats faced by different countries and the irresistible drive to achieve higher economic growth severely hinder genuine attempts to control carbon emissions. As a result, environmental degradation persists, and environmental quality continues to decline significantly (Chu & Lee, 2022; Pradhan et al., 2024). To better understand why environmental quality continues to deteriorate, it is necessary to examine the factors influencing environmental degradation. These factors include, among others, financial development (FD), economic growth (EG), energy consumption (EC), and trade openness (TO) (Khalid et al., 2025; Egbe et al., 2024; Jia et al., 2023).

The development of the financial sector is one of the key prerequisites for sustaining consumption and economic activity. From the perspective of economic activity, FD makes financial resources more available, encourages investment, and can support sectors that rely on high-carbon energy sources such as coal, natural gas, and oil. At the same time, it also has the potential to finance environmentally friendly energy sources. When considering the consumption aspect, FD may lead consumers to increase their spending, which can result in higher demand for energy-intensive products. Moreover, it can also increase demand for low-carbon and energy-efficient goods. In this context, FD can influence environmental degradation in both positive and negative ways (Dridi et al., 2024).

Economic activities consist of production and consumption processes that cannot be considered independently of the environment in which they take place. Accordingly, as the economy grows, its impact on the environment also increases (Kahuthu, 2006, p. 56). The relationship between EG and environmental degradation is often explained through the Environmental Kuznets Curve (EKC) hypothesis. Accordingly, in the early phases of EG, increases in per capita income contribute to a higher level of environmental degradation up to a certain turning point, beyond which further income growth leads to environmental improvement. In general, the EKC indicates an inverted U-shaped relationship between EG and environmental degradation. This turning point in income levels can be rationalized by: (i) the presence of environmental damage resulting from prior EG and resource utilization, (ii) elevated environmental awareness once such damage becomes evident, and (iii) the accessibility of cleaner technologies emerging from previous phases of economic development. While many studies provided empirical support for this inverted U-shaped relationship, others reported mixed, weak, or alternative patterns that better capture the complexities of this relationship (Kostakis & Arauzo-Carod, 2023).

EC is a critical component of EG and must be carefully examined in the context of environmental sustainability. In efforts to stimulate EG, countries consider energy a driving force to sustain production across industries and other productive sectors. Therefore, EC lies at the core of industrialization and EG (Gyamfi et al., 2024). Even though increased EC may initially accelerate economic development, it typically exacerbates environmental degradation by increasing greenhouse gas emissions, primarily due to the combustion of fossil fuels (Rehan et al., 2025). However, the use of conventional fossil fuels continues to dominate the EG strategies of many nations. However, the importance of clean and efficient energy sources in achieving sustainable EG is increasingly being recognized, leading to growing calls for a transition from fossil fuels to clean energy alternatives. Clean energy sources are regarded as non-depletable, less polluting, and environmentally safer. A growing body of literature asserts a positive link between the expansion of clean energy and improvements in environmental quality (Gyamfi et al., 2024).

Trade policy is also considered an important indicator influencing environmental degradation and environmental quality. The impact of TO (i.e., international trade) on environmental degradation is explained through three main aspects. The first one of them is the technological effect, which refers to the enhancement of technological innovation as trade volume increases, subsequently leading to reductions in environmental damage. The second channel, known as the scale effect, suggests that open trade has toxic implications for environmental quality by increasing trade volume and production. Lastly, the composition effect indicates that many less developed economies tend to shift toward sectors in which they have a comparative advantage, thereby attracting pollution-intensive industries that exacerbate environmental degradation. Within this framework, it can be argued that while the technological effect has a direct positive effect on environmental quality, the scale and composition effects tend to reduce it (Usman et al., 2022).

The E7 and G7 countries represent, respectively, some of the world’s fastest-growing and most developed economies. The stable growth of E7 countries has been accompanied by increased EC, which in turn contributes to increasing CO₂ emissions. British Petroleum (2021) estimates that E7 countries alone account for 46% of global carbon emissions. Given this figure, it is projected that the E7 will account for 50% of global GDP by 2050, signaling a major shift in economic power from the G7 to the E7. As these countries accelerate in economic power, their increasing EC results in a higher level of pollutant emissions, exacerbating environmental degradation (Gyamfi et al., 2024).

Given their dominance in the global system and the magnitude of their carbon emissions, this study aims to investigate the relationship among FD, EG, EC, and environmental degradation in E7 (Brazil, China, Indonesia, India, Mexico, Russia, and Türkiye) and G7 (Canada, Germany, France, the United Kingdom, Italy, Japan, and the USA) countries. To this end, a panel cointegration and causality analysis was conducted using annual data from 2000 to 2021 for the E7 and G7 countries. The following points highlighted in this study are expected to contribute to the existing literature: (i) In the current literature, most researchers have selected regional or individual countries as their sample, while only a few (Gyamfi et al., 2022, 2024; U. Khan et al., 2023; Doğan et al., 2022) have comparatively examined the E7 and G7 countries together. Since the literature provides limited evidence on comparative analyses, these country groups were included in the present study. (ii) In studies where the E7 and G7 countries have been examined comparatively, the analysis has been conducted only at the group level, without country-level comparisons. In contrast, this study applies cointegration and causality tests at the country level. (iii) The study provides insights into how variables such as financial development, economic growth, energy consumption, and trade openness can be utilized by countries to reduce CO₂ emissions and protect the environment. (iv) Such comparative studies are important in justifying whether countries at different levels of development require differentiated policies.

After the Section 1, this study reviews the theoretical framework and empirical literature relevant to the topic. It then presents the data and methodology in detail, evaluates the findings obtained from the analysis, and concludes with a discussion and summary of the results.

2. Theoretical Framework and Literature

This section addresses the relationships between FD, EG, EC, TO, and environmental degradation, first through a theoretical perspective and then through the empirical literature.

2.1. FD and Environmental Degradation

Many studies investigated the relationships between FD and environmental degradation using various variables, methodologies, and samples. These studies emphasized that FD affects the environment in multiple ways, through capitalization, technological, income, and regulatory channels (Karl & Chen, 2010).

The capitalization effect posits that capital can be accessed at a lower cost in a developed financial market, thereby facilitating investment and production through adequate capital supply, which in turn may result in an increase in harmful emissions. On one hand, this effect encourages the establishment of small enterprises that often rely on outdated technologies, while on the other, it offers medium- and large-scale firms the opportunity to invest in upgrading their existing production technologies to more environmentally friendly alternatives. Through technological effects, the increased availability of capital can support the development of modern technologies that improve energy efficiency. However, this advancement does not entirely eliminate the risk of resource overuse or the emergence of new pollution sources (such as nuclear power plants). Nevertheless, the technological effect may also facilitate the research and development of modern technologies that are less carbon- and energy-intensive (Chu et al., 2023). The income effect suggests that FD, by enabling greater access to credit, can boost household incomes, allowing consumers to purchase and consume more environmentally friendly products, thereby potentially mitigating environmental degradation. However, from another perspective, rising income levels may increase demand for high-energy-consuming products, which could in turn exacerbate environmental degradation. The regulatory effect indicates that bank credit issued to fund environmentally sustainable investment projects can help reduce environmental degradation. In this context, FD can have both positive and negative effects on environmental degradation through various channels (Lahiani, 2020).

There are many empirical studies at both the country and regional levels demonstrating a positive association between FD and environmental degradation. For example, Boutabba (2014) reported that FD in India had a long-term positive effect on CO₂ emissions. Ahmad et al. (2018), using the NARDL approach for the period 1980–2014, highlighted a positive relationship between FD and CO₂ emissions. Acheampong (2019), in a study on 46 Sub-Saharan African countries, also reported that FD contributes to an increase in CO₂ emissions. Similarly, R. Wang et al. (2020) presented evidence from the N-11 countries for the period 1990–2017, indicating a positive relationship among FD, EG, and CO₂ emissions. Zhang et al. (2022) examined the relationship between FD, renewable energy, digital trade, and ecological footprint in G7 countries over the 2000–2020 period using CUP-FM and CUP-BC estimators. Their findings revealed a positive relationship between FD and ecological footprint, while digital trade and renewable energy were found to have a negative relationship with ecological footprint. Habiba et al. (2023), employing AMG estimators and data from E7 countries between 1990 and 2020, determined that FD increases CO₂ emissions. In a more recent study, Gyamfi et al. (2024) used quantile regression and AMG techniques to analyze data from E7 and G7 countries for the period 1990–2019. Their results show that both FD and EG increase CO₂ emissions in E7 countries, whereas EG increases CO₂ emissions in G7 countries while FD and renewable energy contribute to their reduction.

In the literature, there are also many studies suggesting that FD may have a negative effect on environmental degradation. For instance, Tamazian et al. (2009), in their study on BRIC countries for the period 1992–2004, determined that FD reduces environmental degradation. Shahbaz et al. (2013) employed ARDL and Granger causality analyses on data from 1975 to 2011 to investigate the relationships among EG, FD, EC, TO, and CO₂ emissions in Indonesia. Their findings suggest that while EG and EC increase CO₂ emissions, FD and TO help reduce them. Uddin et al. (2017) examined the effects of EG, FD, and TO on the ecological footprint in 27 high-emission countries between the years 1991 and 2012. Using the DOLS method, their results showed that while EG positively affects the ecological footprint, FD and TO have a negative effect. Similarly, Majeed and Mazhar (2019), using panel data regressions and GMM estimations on data from 131 countries for the 1971–2017 period, found that FD reduces ecological footprint, whereas EG and EC increase it. Egbe et al. (2024) employed DFE-ARDL and PMG-ARDL methods to analyze data from G7 countries for the period between 1990 and 2020, focusing on the effects of FD, EG, renewable energy, population growth, and globalization on CO₂ emissions. Their results indicate that renewable energy, globalization, and FD mitigate environmental degradation, whereas EG and population increase have short-term adverse effects but contribute to environmental improvements in the long run. In a study carried out by Degirmenci et al. (2025), the effects of FD, human development, urbanization, and industrial employment on environmental quality in E7 countries for the period 1991–2019 were analyzed using the Durbin-Hausman cointegration test along with CCE and AMG estimators. The results indicate that FD enhances environmental quality in Russia and India, human development increases environmental pollution in China, urbanization improves environmental quality in Brazil but degrades it in China, and industrial employment increases environmental quality in Brazil, China, Indonesia, and India. Finally, Rehan et al. (2025) investigated the relationships among renewable energy, FD, trade, and CO₂ emissions in G7 and BRICS countries, and they reported that FD and renewable energy reduce CO₂ emissions, whereas trade increases them.

Studies reporting no relationship between FD and environmental degradation or finding no significant impact of FD on environmental degradation include, for instance, the study carried out by Ziaei (2015), who applied Panel Vector Autoregression (PVAR) models to data from 13 European and 12 East Asian and Oceanian countries for the period 1989–2011. The findings of this study indicated that FD had no significant effect on CO₂ emissions. Similarly, Charfeddine and Kahia (2019), using PVAR models and impulse response analyses for data covering 24 countries in the Middle East and North Africa (MENA) region over the 1980–2015 period, found that both renewable energy consumption and FD had insignificant impacts on CO₂ emissions. The study carried out by Opuala et al. (2023), using PMG and AMG estimators on data from West African countries for the period 1980–2017, also revealed that FD and natural resource rents do not make a substantial contribution to environmental quality. However, income, EC, TO, and urbanization were found to significantly exacerbate environmental degradation.

2.2. EG and Environmental Degradation

The relationship between EG and environmental degradation is primarily examined within the framework of the EKC hypothesis (Çetin et al., 2023). This concept was first introduced in 1955 by economist Simon Kuznets to describe the relationship between EG and income inequality. He suggested that income inequality tends to increase in the early phases of industrialization, but it reaches a peak and then declines as economies mature. Grossman and Krueger (1991) later extended this hypothesis to environmental economics, arguing that environmental degradation initially gets worse with EG but improves after reaching a certain income threshold. The EKC asserts an inverted U-shaped relationship between per capita emissions and per capita GDP. In the early phases of EG, industrialization and expanding economic activities generally yield higher pollution levels. However, during the further development of economies, investments in cleaner technologies and stricter environmental regulations contribute to the mitigation of environmental degradation (Khalid et al., 2025).

The literature on the relationship between EG and environmental degradation reported mixed findings. Many studies support the EKC hypothesis, indicating an inverted U-shaped relationship between EG and environmental degradation. For example, Galeotti et al. (2006) reported evidence of such an association in OECD countries. Dutt (2009) confirmed the EKC hypothesis for 124 countries for the 1982–2002 period. Bölük and Mert (2015), analyzing time series data for Türkiye for a period between 1961 and 2010, also validated the EKC hypothesis. P. Y. Chen et al. (2016) conducted a panel data analysis covering 188 countries for the 1993–2010 period and concluded that CO₂ emissions and EG have an inverted U-shaped relationship. Aydoğan and Vardar (2019) found supporting evidence for the EKC utilizing panel data for the E7 countries over the 1990–2014 period. Gyamfi et al. (2021), employing PMG-ARDL models for E7 countries using 1995–2018 data, demonstrated an inverted U-shaped relationship between EG and CO₂ emissions. Husnain et al. (2022) conducted AMG and panel causality analyses to data from the E7 countries for the 1990–2015 period and found evidence consistent with the EKC hypothesis. Ahmad et al. (2023), using various panel data models for the 1990–2018 period in E7 countries, also reported findings that support the EKC hypothesis.

Nonetheless, some studies in the literature report alternative forms of the relationship between EG and environmental degradation, such as N-shaped or other nonlinear patterns. For instance, Friedl and Getzner (2003), in their study on Austria for the 1960–1999 period, identified an N-shaped relationship between EG and CO₂ emissions. Martınez-Zarzoso and Bengochea-Morancho (2004) found a similar N-shaped relationship in 22 OECD countries for the 1975–1998 period. Poudel et al. (2009), analyzing data from 15 Latin American countries for the period between 1980 and 2000, also identified a relationship resembling an N shape. Özokçu and Özdemir (2017), through panel data analysis of 26 OECD and 52 developing countries for the 1980–2010 period, reported evidence of both N-shaped and inverted N-shaped relationships between EG and CO₂ emissions. Allard et al. (2018) found evidence of an N-shaped EKC in all income groups, except for upper-middle-income countries, using data from 74 countries for the 1994–2012 period. Doğan et al. (2022), using 1991–2017 data for G7 and E7 countries, found an inverted U-shaped relationship between economic complexity and carbon emissions in the G7 countries, while a U-shaped relationship was identified in the E7 countries. Lastly, Fakher et al. (2023), using the DOLS method on data from OPEC countries for the 1994–2019 period, revealed an N-shaped linkage between EG and environmental degradation.

In the literature, there are also studies suggesting a causal relationship between EG and environmental degradation. For instance, panel cointegration and causality analyses conducted by Dritsaki and Dritsaki (2014) for Southern European countries covering the period 1960–2009 revealed a short-run bidirectional causal relationship between EG and CO₂ emissions, and a unidirectional causality from CO₂ emissions to EG in the long run. Similarly, Saidi and Hammami (2017), using data from 75 countries for a period between 2000 and 2014, found evidence supporting a bidirectional causality between EG and environmental degradation. The results of Emirmahmutoğlu and Köse (2011) causality test applied by Yıldız (2019) to E7 countries for the period 1992–2004 indicate a unidirectional causality from EG to CO₂ emissions in Russia, Brazil, and India, and from CO₂ emissions to EG in China. Tong et al. (2020), employing ARDL and Granger causality analyses for E7 countries over the period 1971–2014, found evidence of short-run causality from both EC and EG to CO₂ emissions in all E7 countries except Indonesia. Panel cointegration and causality analyses conducted by Murshed et al. (2022) using data from G7 countries spanning 1995–2016 demonstrated that EG exacerbates environmental degradation and that there exists a bidirectional causal relationship between EG and environmental degradation. Elhassan (2025), employing panel cointegration and causality tests for the G7 countries for the period 1990–2022, determined that EG positively influences CO₂ emissions but negatively affects the ecological footprint. The findings also indicate a unidirectional causal relationship running from EG to both CO₂ emissions and ecological footprint.

2.3. EC and Environmental Degradation

With the growth of population, increasing income levels, and industrial development, energy demand and consumption are steadily increasing. The sources directly used to meet energy needs are referred to as primary energy sources. These are energy types used in their natural form, without requiring processing or conversion. Primary energy sources include non-renewable fossil fuels (coal, oil, and natural gas) and renewable energy sources such as hydropower, solar, wind, and geothermal energy. Although fossil fuels still constitute a significant portion of today’s global EC, growing environmental awareness and sustainability concerns have led to increased interest in renewable energy sources (İnanç, 2024).

In the early stages of EG, energy needs are primarily met through fossil fuels. However, the use of fossil fuels significantly increases the emission of CO₂ and other harmful pollutants. In addition to depleting natural resources, fossil fuel consumption raises concerns about air and water pollution. The emissions generated by the increased use of fossil fuels lead to atmospheric and ecosystem degradation, thereby negatively impacting environmental quality (Kartal et al., 2025).

Many empirical studies in the literature addressed the relationship between EC and environmental degradation, generally reporting that the use of non-renewable fossil fuels decreases environmental quality. For example, Ozturk and Acaravcı (2013), who used data from Türkiye covering the period 1960–2007, investigated the relationships among FD, trade, EG, EC, and carbon emissions using cointegration, bounds, and causality tests. The cointegration test results indicated that trade increases carbon emissions, while FD does not have a significant effect on emissions. Nevertheless, evidence was found of long-run causality from FD, EC, EG, and TO to CO₂ emissions.

Sehrawat et al. (2015), examining time series data from India for the period 1971–2011, reported that EC, EG, FD, and urbanization all contributed to increased environmental degradation. Doğan and Değer (2018), through panel cointegration analyses using 1990–2014 data from E7 countries, concluded that EC increases carbon emissions. Tong et al. (2020), employing the Bootstrap ARDL bounds testing and panel causality analysis for E7 countries for the period 1971–2014, found evidence of a short-run bidirectional causality between CO₂ emissions and EC in all E7 countries.

Pradhan et al. (2024), analyzing South Asian and G7 countries using the ARDL method for the period 1996–2021, reported that EC significantly increases CO₂ emission levels. Liang et al. (2024), using NLPARDL analysis on data from E7 countries for the period of 1995–2022, showed that positive shocks to EC lead to higher levels of increase in CO₂ emissions over the long term. Lastly, Elhassan (2025), employing CS-ARDL and Dumitrescu-Hurlin (2012) panel causality analyses for G7 countries between 1990 and 2022, concluded that EC positively affects both CO₂ emissions and the ecological footprint.

Empirical literature also supports the view that investing in renewable energy infrastructure aiming to reduce fossil fuel consumption is a critical strategy for mitigating environmental damage (Khalid et al., 2025). There are studies demonstrating that renewable EC helps slow environmental degradation. For instance, Jebli et al. (2020), using the System Generalized Method of Moments (System-GMM) and Granger causality tests for the period of 1990–2015 across 102 countries, found that renewable EC reduces CO₂ emissions in all countries except for those classified as lower-middle-income. Similarly, Safi et al. (2021), employing CS-ARDL, CCEMG, and AMG models for the G7 countries for the 1990–2018 period, revealed that renewable energy significantly reduces carbon emissions. Lu (2022), applying CUP-FM, CUP-BC, and the panel causality test developed by Dumitrescu and Hurlin (2012) for E7 countries for the period 1995–2018, also concluded that renewable energy positively contributes to the reduction of CO₂ emissions. However, Murshed et al. (2022), using AMG and Dumitrescu-Hurlin panel causality analysis for G7 countries between 1995 and 2016, reported findings indicating that renewable EC worsens environmental degradation in those countries. In a similar vein, Gyamfi et al. (2022), utilizing FMOLS, DOLS, and AMG methods for G7 countries during the 1990–2016 period, presented evidence that renewable energy negatively affects CO₂ emissions. In contrast, Kostakis and Arauzo-Carod (2023), using PMG-ARDL and Dumitrescu-Hurlin panel causality analyses with data from 1980–2018 for G7 countries, demonstrated that renewable energy improves environmental sustainability. U. Khan et al. (2023), through CS-ARDL, DOLS, and Dumitrescu-Hurlin causality models for the period 1997–2018 for both E7 and G7 countries, found evidence suggesting that renewable energy improves environmental quality. Supporting this, Gyamfi et al. (2024) employed cointegration analyses for E7 and G7 countries during 1990–2019 and found that renewable energy contributes to reducing CO₂ emissions. Egbe et al. (2024), using DFE-ARDL and PMG-ARDL methods for G7 countries for the period between 1990 and 2020, concluded that renewable energy yields environmental benefits. Finally, H. Chen et al. (2025) demonstrated through econometric analyses of G7 data from 1990 to 2022 that renewable energy reduces CO₂ emissions, and Rehan et al. (2025) reported similar findings for G7 and BRICS countries.

Moreover, the literature also includes studies examining the combined effects of renewable and non-renewable EC on environmental degradation. Aydoğan and Vardar (2019), employing panel cointegration and causality analyses for E7 countries for the period 1990–2014, found that non-renewable EC increases CO₂ emissions, whereas renewable EC decreases them. They also reported a bidirectional causality between non-renewable EC and CO₂ emissions. Danish et al. (2020), using FMOLS and DOLS methods for BRICS countries for the period between 1992 and 2016, reported that renewable EC reduces the ecological footprint. Usman et al. (2021), through cointegration analysis covering the top 15 CO₂-emitting countries for the period of 1990–2017, revealed that FD, renewable energy, and TO decelerate environmental degradation, while EG and non-renewable energy usage accelerate it. Their panel causality analysis also indicated bidirectional causality between FD, EG, renewable energy use, and the ecological footprint, as well as unidirectional causality from non-renewable energy and TO to the ecological footprint. Finally, Husnain et al. (2022), employing AMG and Dumitrescu-Hurlin panel causality tests with data from E7 countries for the period of 1990–2015, found that renewable energy reduces both the ecological footprint and CO₂ emissions, whereas non-renewable EC increases both.

2.4. TO and Environmental Degradation

The relationship between TO and environmental degradation is commonly explained in the literature through several channels, including the scale effect, the technological effect, and the composition effect. The scale effect suggests that increased trade among countries leads to higher production in pursuit of higher income, which in turn deteriorates environmental quality because of increased output. The technological effect, by contrast, emphasizes that the use of existing technologies and the development of cleaner technologies can mitigate environmental degradation. The composition effect refers to the tendency of countries to shift toward pollution-intensive production and to import inexpensive, pollution-intensive technologies that support such production, thereby enhancing EG while diminishing environmental quality. In this context, the scale and composition effects are generally understood to exacerbate environmental degradation, whereas the technological effect is thought to slow it down (Destek & Sinha, 2020; Çetin et al., 2022).

The relationship between TO and environmental degradation has also been examined within the scope of two main theoretical frameworks. The first is the pollution haven hypothesis, which posits that TO adversely affects the environment (Kartal et al., 2025). Numerous empirical studies have demonstrated that TO decreases environmental quality and intensifies environmental degradation. For example, Tamazian and Rao (2010) found, using GMM estimation on data from 24 transition economies for the period 1993–2004, that TO increases environmental degradation. Shahbaz et al. (2017) similarly concluded, in a study on 105 countries, that TO diminishes environmental quality. In Türkiye, Pata (2019) analyzed data for the period of 1969–2017 and found that TO increases CO₂ emissions and that there is a unidirectional causal relationship from TO to CO₂ emissions. Mikayilov et al. (2019), analyzing data for the period of 1996–2014 for Azerbaijan, found that trade positively affects the ecological footprint. Ibrahim and Ajide (2021) presented evidence that TO increases CO₂ emissions in G7 countries between 1990 and 2019. Akhayere et al. (2023) used quantile regression analysis for Türkiye for the period of 1965–2018 and found evidence that TO adversely affects environmental quality. Thi et al. (2023), employing FMOLS, D2OLS, and GMM analyses on data from 53 countries for the period 1990–2019, also found that TO contributes to environmental degradation.

In contrast, the second theory, known as the beneficial trade hypothesis, asserts that TO can improve environmental quality (Kartal et al., 2025). There are studies supporting this view. For instance, Managi et al. (2009) concluded that trade benefits the environment in a study conducted on OECD countries. Liu et al. (2022) conducted a cointegration analysis on data from Pakistan for the period between 1980 and 2017 and found that trade has a positive effect on environmental quality. Kartal et al. (2025) also provided evidence, based on CS-ARDL and CCEMG analyses for E7 countries for the period of 1990–2022, that TO enhances environmental quality.

Beyond these two dominant theories, there are also empirical studies that report heterogeneous findings, such as the absence of any relationship between the two variables, the coexistence of both positive and negative relationships, and the presence of causality. For example, Kasman and Duman (2015), using data for the period between 1992 and 2010 for a panel of EU member and candidate countries, found a short-run unidirectional causality from TO to carbon emissions and a long-run effect of TO on carbon emissions. Le et al. (2016), using data from 98 countries between 1990 and 2013, reported that TO has a positive effect on the environment in high-income countries, while it has a negative effect in middle- and low-income countries. Tachie et al. (2020) reported evidence of unidirectional causality from TO to CO₂ emissions in 18 EU countries. Essandoh et al. (2020), using PMG-ARDL analysis for 52 developing and developed countries for the period of 1991–2014, found that CO₂ emissions are negatively associated with trade in developed countries but positively associated in developing countries in the long run. Similarly, H. Khan et al. (2021) showed that for the 1985–2018 period, TO reduces carbon emissions in developed countries but decreases environmental quality in developing ones. Q. Wang et al. (2024), analyzing data from 96 developing countries for the 2000–2018 period, identified a threshold effect of TO on environmental quality, indicating a U-shaped relationship. Their findings also suggest that TO improves environmental quality at higher levels of natural resource rents and corruption control. Pham and Nguyen (2024), in their study of 64 developing countries using data for the period of 2003–2017, found no significant effect of TO on environmental pollution. Finally, Barkat et al. (2024), in their study of 20 OECD countries, found that TO directly increases CO₂ emissions but indirectly reduces them through the income growth channel.

Considering the results of the reviewed studies within the literature and the purpose of this research, the hypotheses can be expressed as follows:

H1a: 

Financial development positively affects CO₂ emissions.

H1b: 

Financial development negatively affects CO₂ emissions.

H1c: 

Financial development has no effect on CO₂ emissions.

H2a: 

Economic growth positively affects CO₂ emissions.

H2b: 

Economic growth negatively affects CO₂ emissions.

H2c: 

Economic growth has no effect on CO₂ emissions.

H3a: 

Energy consumption positively affects CO₂ emissions.

H3b: 

Energy consumption negatively affects CO₂ emissions.

H3c: 

Energy consumption has no effect on CO₂ emissions.

H4a: 

Trade openness positively affects CO₂ emissions.

H4b: 

Trade openness negatively affects CO₂ emissions.

H4c: 

Trade openness has no effect on CO₂ emissions.

3. Data

Within the scope of this study, the relationship between FD, EC, EG, TO, and environmental degradation is examined in E7 countries (Brazil, China, Indonesia, India, Mexico, Russia, and Türkiye) and G7 countries (Canada, Germany, France, the United Kingdom, Italy, Japan, and the USA). Based on considerations of data availability and continuity, the analysis was conducted using annual data covering the period of 2000–2021. In this study, the FD variable is represented by the FD index, the EC variable by primary energy consumption, the EG variable by per capita GDP growth, the environmental degradation variable (CO₂) by carbon dioxide emissions, and the TO variable by the ratio of total exports and imports to GDP. The Financial Development Index is a measure that ranks countries relative to one another based on the depth, access, and efficiency of their financial markets and institutions. The Financial Development Index is composed of the aggregate of the Financial Institutions Index and the Financial Markets Index (IMF, 2025). Representing financial development in a multidimensional manner, the Financial Development Index is frequently employed in the literature (Balsalobre-Lorente et al., 2023; Habiba et al., 2023; Zhang et al., 2022; Degirmenci et al., 2025; Rehan et al., 2025; Opuala et al., 2023). The data used in the analysis were obtained from the official websites of the World Bank, IMF and Our World in Data (World Bank, 2025; IMF, 2025; Our World in Data, 2025a, 2025b). Before proceeding with the econometric analysis, the natural logarithms of all variables except FD and EG were taken. Descriptive statistics for the variables used in the econometric analysis are presented in

Table 1. Descriptive statistics for the variables.

E7 Countries
	CO2	EG	EC	FD	TO
Average	8.966	3.719	3.606	0.457	1.659
Maximum	10.059	4.129	4.642	0.670	1.919
Minimum	8.330	2.879	2.895	0.270	1.345
Std. Dev.	0.473	0.334	0.442	0.100	0.130
Jarque–Bera	21.002	25.125	13.719	4.885	9.558
Probability	0.000	0.000	0.001	0.087	0.008
Observation	154	154	154	154	154
G7 Countries
Average	8.907	4.597	3.641	0.818	1.689
Maximum	9.788	4.799	4.432	0.960	1.918
Minimum	8.448	4.469	3.218	0.670	1.291
Std. Dev.	0.387	0.076	0.349	0.076	0.168
Jarque–Bera	42.860	10.532	51.457	11.336	20.106
Probability	0.000	0.005	0.000	0.003	0.000
Observation	154	154	154	154	154

.

Table 1 reveals that, among the E7 countries, CO₂ has the highest mean and standard deviation, while FD has the lowest mean and standard deviation. Moreover, the results of the Jarque–Bera test confirm that, at the 5% significance level, all variables except for FD in the E7 countries and all variables in the G7 countries do not have a normal distribution. One of the most important theorems in statistics, the Central Limit Theorem, asserts that for most non-normal distributions, as the sample size increases, the distribution approaches normality (Hays, 1994). Gujarati and Porter (2009) note that the normality assumption can be relaxed if the sample size is reasonably large. In addition, several studies (Uttley, 2019; Büyüköztürk et al., 2020) emphasize that when the sample size exceeds 30, the distribution tends to approximate a normal distribution as the number of observations grows. In light of this information, the econometric analyses in this study were conducted under the assumption that the data employed (154 observations) follow a normal distribution. This study includes 154 observations for each of the E7 and G7 countries and uses two balanced panel datasets: one for the E7 countries and one for the G7 countries, each comprising 7 countries, a 22-year period, and 154 observations. The model developed for both panels is represented as follows:

$${CO2}_{it}={\alpha }_{it}+{\beta }_1{FD}_{it}+{{\beta }_{2}EG}_{it}+{{\beta }_{3}EC}_{it}+{{\beta }_{4}TO}_{it}+u_{it} i=\mathrm{1,2}\dots .7 t=\mathrm{1,2},\dots 22$$

(1)

In Equation (1), CO₂ represents the environmental degradation variable, FD the financial development variable, EG the economic growth variable, EC the energy consumption variable, and TO the trade openness variable.

4. Methodology

The data used in the analysis consist of both cross-sectional (e.g., countries, firms) and time-series components. Such data structures are referred to as panel data. To ensure the use of appropriate methods in panel data analyses, a series of preliminary tests must be conducted. The first of these are cross-sectional dependence tests. Based on the results of these tests, panel unit root tests that either account for or ignore cross-sectional dependence are then employed. In addition to these preliminary tests for the variables, further diagnostics are also performed on the models to determine the presence of cross-sectional dependence and parameter homogeneity. Depending on the results of these diagnostics, various panel cointegration and causality tests are applied to analyze the relationships among variables. Accordingly, this study employs cross-sectional dependence, panel unit root, homogeneity, panel cointegration, and panel causality analyses as its methodological framework.

4.1. Cross-Sectional Dependence Tests

Cross-sectional dependence refers to the degree to which units within the cross-section (such as countries) respond similarly to a common shock in the global system. If the cross-sectional units are affected to the same degree by a shock, then there is no dependence. If the degree of impact varies across units, then cross-sectional dependence is present (Degirmenci & Aydın, 2021, p. 2234). Given the rising levels of globalization, international trade, and financial integration, the potential for shocks in global markets to affect multiple countries underscores the importance of accounting for cross-sectional dependence. Pesaran (2007) highlights the significance of addressing cross-sectional dependence in panel data analyses, showing that neglecting it can lead to substantial biases and distortions in the results (Kar et al., 2011, p. 688).

Several tests are used to detect cross-sectional dependence, including the LM test developed by Breusch and Pagan (1980), the CD_LM and CD tests proposed by Pesaran (2004), and the LM_adj test introduced by Pesaran et al. (2008). The LM and CD_LM tests tend to yield substantial distortions in experimental applications when the number of cross-sectional units (N) is large, and the time dimension (T) is small. To address this issue, Pesaran (2004) introduced the CD test, which is applicable when N > T. The CD test is considered valid as long as either N or T tends to infinity. The LM_adj test, on the other hand, compensates for the shortcomings of the LM test regarding the number of cross-sectional units and provides consistent results even when the CD test is unreliable. Therefore, this study employs the LM_adj test. The test statistic is formulated as follows (Pesaran, 2004; Pesaran et al., 2008):

$$L{M}_{adj}=\sqrt{{\displaystyle \frac{2}{N(N-1)}}}{\sum }_{i=1}^{N-1}{\sum }_{j=i+1}^N{\displaystyle \frac{\left(T-k\right){\rho }_{ij}^2-{\mu }_{Tij}}{{\upsilon }_{Tij}}}\to N(0,1)$$

(2)

In Equation (2), $\widehat{\rho }$ denotes the correlation coefficient between error terms, k represents the number of explanatory variables, N the number of cross-sectional units, and T the number of time periods. μ_Tij and υ_Tij denote the exact mean and variance of $\left(T-k\right){\rho }_{ij}^2$, respectively. Under the null hypothesis of no cross-sectional dependence, as N→∞ and T→∞, the LM_adj → _dN(0,1) (Pesaran et al., 2008, p. 108).

This test eliminates the bias in the LM test and the possibility that the correlation sum in the CD test equals zero, and it is used when the time dimension (T) exceeds the cross-sectional dimension (N) (Balsalobre-Lorente et al., 2023). Due to its advantages, the LMadj test is frequently applied in the literature (Ehigiamusoe & Dogan, 2022; R. Li et al., 2023; Hossain et al., 2024; H. Khan et al., 2021; Destek & Sarkodie, 2019; Le et al., 2016; Cowan et al., 2014).

4.2. CIPS Panel Unit Root Test

In the presence of cross-sectional dependence, panel unit root tests that account for this dependence must be utilized. The Cross-Sectionally Augmented IPS (CIPS) test is one of such second-generation panel unit root tests. Developed by Pesaran (2007), the CIPS test is expressed as follows:

$$CIPS\left(N,T\right)= N^{-1}\sum _{i=1}^N{t}_i(N,T)$$

(3)

In Equation (3), t_i(N,T) refers to the CADF (Cross-Sectionally Augmented Dickey–Fuller) test statistic for the ith cross-sectional unit. The CIPS test statistic is calculated by averaging the CADF statistics across all units. The null hypothesis of the CIPS test posits that the series contain a unit root. The test statistic is then compared to the critical values obtained from Monte Carlo simulations introduced by Pesaran (2007). If the test statistic is lower than the critical value, the null hypothesis is rejected, suggesting that the series are stationary (Degirmenci & Aydın, 2021, p. 2234).

The CIPS test, by accounting for cross-sectional dependence and slope heterogeneity, produces more consistent and reliable results (Pesaran, 2007). For this reason, it is widely employed in the literature (Balsalobre-Lorente et al., 2023; R. Li et al., 2024; Jahanger et al., 2022; Chien et al., 2021; R. Wang et al., 2020; Nathaniel & Iheonu, 2019; Dogan & Seker, 2016).

4.3. Cross-Sectional Dependence and Homogeneity Tests of the Model

In panel data analyses based on the established model, it is necessary to examine the model in terms of cross-sectional dependence and parameter homogeneity to determine the appropriate estimation methods. The methods must be compatible with the characteristics of the data. Tests for cross-sectional dependence, discussed in the relevant section, can be applied to both individual series and models. Hence, these tests will not be revisited here. In panel data models, the homogeneity of slope coefficients can be assessed using the Delta (∆) and adjusted Delta (∆_adj) tests developed by Pesaran and Yamagata (2008). The formulations for these tests are presented in Equations (4) and (5) (Pesaran & Yamagata, 2008, p. 57):

$$\widetilde{∆}=\sqrt{N}\left({\displaystyle \frac{N^{-1}\tilde{S}-k}{\sqrt{2k}}}\right)$$

(4)

$${\widetilde{∆}}_{adj}=\sqrt{N}\left({\displaystyle \frac{N^{-1}\tilde{S}-E\left({\tilde{z}}_{iT}\right)}{\sqrt{Var\left({\tilde{z}}_{iT}\right)}}}\right)$$

(5)

The tests in Equations (4) and (5) are standardized versions of Swamy’s (1970) homogeneity test. The null hypothesis of these tests states that the slope coefficients are homogeneous. Generally, the ∆ test is recommended for large samples, whereas the ∆_adj test is more suitable for small samples (Pesaran & Yamagata, 2008, p. 57).

The homogeneity test is particularly important for verifying whether slope coefficients remain constant across observations and over time. This test is commonly applied in panel regression analysis to determine whether heterogeneous coefficients exist that may ensure the consistency and accuracy of parameter values in the model, and it has been used extensively in the literature (Adebayo et al., 2023; Dai et al., 2025; G. Wang et al., 2022; Udeagha & Ngepah, 2023; Rahman et al., 2024; Murshed, 2021; Dogru & Bulut, 2018; Halliru et al., 2020).

4.4. LM Bootstrap Panel Cointegration Test

The cointegration test is a critical step carried out after unit root, homogeneity, and cross-sectional dependence tests. This test determines whether there exists a long-run relationship among the variables under investigation (Ahmad et al., 2023). To investigate the relationships among the variables in the E7 and G7 countries, the LM bootstrap cointegration test was employed. Developed by Westerlund and Edgerton (2007), this test considers cross-sectional dependence and yields robust results in small samples. The test statistic is calculated using the formulation in Equation (6) (Westerlund & Edgerton, 2007, p. 186):

$${LM}_N^{+}={\displaystyle \frac{1}{{NT}^2}}\sum _{i=1}^N\sum _{t=1}^T{\widehat{\omega }}_i^{-2}{S}_{it}^2$$

(6)

In Equation (6), ${\widehat{\omega }}_i^{-2}$ denotes the long-run variance of the error term, while S_it² represents the partial sum of the error term. The null hypothesis of the LM test suggests the existence of cointegration among the variables (Westerlund & Edgerton, 2007, p. 186).

Capable of being applied under conditions of both the presence and absence of cross-sectional dependence, and suitable for small samples, this test (Westerlund & Edgerton, 2007) is frequently used in the literature (Afonso & Rault, 2010; Cho et al., 2013; Chortareas et al., 2015; Damette & Marques, 2019; Degirmenci & Aydın, 2021; Koseoglu et al., 2022; Shahbaz et al., 2023).

4.5. Panel AMG Cointegration Estimator

After confirming cointegration among the variables, it becomes necessary to estimate the model that includes the long-run coefficients. In this study, long-run coefficients are estimated using the Panel AMG (Augmented Mean Group) estimator, which accounts for cross-sectional dependence. Developed by Eberhardt and Bond (2009), the AMG estimator calculates the average group effect by weighting individual slope coefficients. The process is conducted in two steps (Eberhardt & Bond, 2009, p. 3).

$$\begin{gathered} Step 1 { ∆y}_{it}=b_i^{\prime }∆x_{it}+\sum _{t=2}^T{c}_t∆D_t+e_{it} \\ {\widehat{c}}_t\equiv {\widehat{\mu }}_t^{∎} \end{gathered}$$

(7)

$$\begin{gathered} Stage 2{ y}_{it}=a_{it}+b_i^{\prime }{x}_{it}+c_{i}t+{d_i\widehat{\mu }}_t^{∎}+e_{it} \\ {\widehat{b}}_{AMG}=N^{-1}\sum _{i=1}^N{\widehat{b}}_i \end{gathered}$$

(8)

In the first stage, a regression model is formed using first differences and T−1 time dummies. The coefficients of these dummies are then aggregated and redefined as ${\widehat{\mu }}_t^{∎}$. This term is included in the second-stage regression for each of N-number cross-sectional units. The AMG estimator is computed as the average of the individual coefficients obtained from these regressions (Eberhardt & Bond, 2009, p. 3).

By considering the issue of cross-sectional dependence and allowing for heterogeneous slope coefficients across units, this test estimates long-run parameters. Moreover, in the test, the unobserved common factors in the AMG technique are considered as part of a shared dynamic process. Therefore, findings derived from the AMG estimator are more robust and reliable compared to other methods (Paramati et al., 2022). Owing to these advantages, it is widely adopted in the literature (Saqib et al., 2024; Bashir et al., 2025; R. Wang et al., 2023; Rafique et al., 2020; Sadorsky, 2013; Geng & He, 2021; Shah et al., 2022; Aydin, 2019).

4.6. Panel Causality Test

After the cointegration analysis, the causal relationships among variables were examined using the bootstrap panel causality test developed by Kónya (2006), which conducts Granger causality testing for each cross-sectional unit in panel data (Kónya, 2006, pp. 990–991). Granger causality refers to a situation in which the current value of one variable can be explained by the past values of another (Brooks, 2014, p. 335). In other words, it measures the predictive power of one variable over another, rather than capturing short-run dynamic responses (Akkuş, 2021, p. 283).

In panel causality analysis, cross-section-specific bootstrap critical values are derived using the SUR (Seemingly Unrelated Regression) estimation method and the Wald test. If the Wald statistics exceed the corresponding critical values, the null hypothesis of no causality is rejected, indicating a causal relationship (Kónya, 2006, pp. 985–991). Furthermore, if the p-values associated with the Wald tests are significant at conventional levels (1%, 5%, or 10%), causality is inferred. This test is considered more practical than other panel causality tests because it does not require prior unit root or cointegration testing. Additionally, it accounts for cross-sectional dependence and slope heterogeneity inherent in panel datasets. The system of equations for the bootstrap panel causality test is specified as follows (Kónya, 2006, p. 981):

$$y_{1,t}=a_{\mathrm{1,1}}+\sum _{i=1}^{mly1}{\beta }_{\mathrm{1,1},i}{y}_{1,t-i}+\sum _{i=1}^{mlx1}{\gamma }_{\mathrm{1,1},i}{x}_{1,t-i}+{\epsilon }_{\mathrm{1,1},t}$$

(9)

$$y_{2,t}=a_{\mathrm{1,2}}+\sum _{i=1}^{mly1}{\beta }_{\mathrm{1,2},i}{y}_{2,t-i}+\sum _{i=1}^{mlx1}{\gamma }_{\mathrm{1,2},i}{x}_{2,t-i}+{\epsilon }_{\mathrm{1,2},t}$$

(10)

$$⋮$$

$$y_{N,t}=a_{1,N}+\sum _{i=1}^{mly1}{\beta }_{1,N,i}{y}_{N,t-i}+\sum _{i=1}^{mlx1}{\gamma }_{1,N,i}{x}_{N,t-i}+{\epsilon }_{1,N,t}$$

(11)

$$\mathrm{and}$$

$$x_{1,t}=a_{\mathrm{2,1}}+\sum _{i=1}^{mly2}{\beta }_{\mathrm{2,1},i}{y}_{1,t-i}+\sum _{i=1}^{mlx2}{\gamma }_{\mathrm{2,1},i}{x}_{1,t-i}+{\epsilon }_{\mathrm{2,1},t}$$

(12)

$$x_{2,t}=a_{\mathrm{2,2}}+\sum _{i=1}^{mly2}{\beta }_{\mathrm{2,2},i}{y}_{2,t-i}+\sum _{i=1}^{mlx2}{\gamma }_{\mathrm{2,2},i}{x}_{2,t-i}+{\epsilon }_{\mathrm{2,2},t}$$

(13)

$$⋮$$

$$x_{N,t}=a_{2,N}+\sum _{i=1}^{mly2}{\beta }_{2,N,i}{y}_{N,t-i}+\sum _{i=1}^{mlx2}{\gamma }_{2,N,i}{x}_{N,t-i}+{\epsilon }_{2,N,t}$$

(14)

In the equation set presented above, y refers to the dependent variable, x to the independent variables, l to the lag length, N to the number of cross-sectional units, i to the cross-sectional units, and t to the time dimension. The null hypothesis of the bootstrap panel causality test asserts that there is no causal relationship running from variable x to variable y (under the restriction that γ_1,i,l = 0 for all i and l) (Kónya, 2006, p. 985).

This method offers several advantages, such as identifying group characteristics and behaviors without requiring pre-tests like unit root or cointegration tests, accounting for cross-sectional dependence and cross-country heterogeneity, and detecting causal relationships among multiple panel members (Kónya, 2006). Consequently, it has been widely employed in the literature (Chou, 2013; Lee et al., 2020; Menyah et al., 2014; Jin & Kim, 2018; U. Khan et al., 2023; Khurshid et al., 2024; Xinyu et al., 2025).

5. Findings

Before examining the relationships among the variables in the panel data set comprising the E7 and G7 countries, it is necessary to conduct a series of preliminary tests. The first of these is the cross-sectional dependence test. The results of this test, applied to the variables corresponding to the E7 and G7 countries, are presented in

Table 2. Cross-sectional dependence test results.

E7 Countries
	CO2		EG		EC		FD		TO
Test	Stat.	Prob.	Stat.	Prob.	Stat.	Prob.	Stat.	Prob.	Stat.	Prob.
LMadj	47.487	0.000 ***	47.380	0.000 ***	58.940	0.000 ***	26.969	0.000 ***	17.242	0.000 ***
G7 Countries
	CO2		EG		EC		FD		TO
Test	Stat.	p	Stat.	p	Stat.	p	Stat.	p	Stat.	p
LMadj	42.874	0.000 ***	45.514	0.000 ***	37.923	0.000 ***	14.673	0.000 ***	33.017	0.000 ***

Note: *** Significant at the 1% level.

.

As seen in Table 2, the p-values associated with all variables in both country groups are significant at the 1% level. This finding leads to the rejection of the null hypothesis of no cross-sectional dependence, indicating that the variables exhibit cross-sectional dependence. In this context, it can be concluded that countries within the E7 and G7 groups are influenced by shocks or fluctuations occurring in other countries within their respective groups.

Following the cross-sectional dependence tests, the next step is to perform panel unit root tests, which are another set of preliminary analyses. In this study, the CIPS test, one of the second-generation panel unit root tests that consider cross-sectional dependence, was applied, and the results are presented in

Table 3. CIPS panel unit root test results.

		E7 Countries		G7 Countries
Variable		Level	First Difference	Level	First Difference
CO2	Constant	−2.052	−2.661 ***	−2.080	−3.512 ***
	Constant + trend	−2.054	−2.774 *	−1.946	−3.548 ***
EG	Constant	−1.567	−2.289 *	−2.264 *	−3.537 ***
	Constant + trend	−1.265	−2.754 *	−2.362	−3.481 ***
EC	Constant	−2.055	−2.815 ***	−1.668	−3.799 ***
	Constant + trend	−2.933 **	−3.141 **	−2.451	−3.935 ***
FD	Constant	−2.350 **	−3.809 ***	−1.777	−3.913 ***
	Constant + trend	−2.394	−3.971 ***	−2.644	−3.808 ***
TO	Constant	−0.982	−2.777 ***	−1.487	−2.424 **
	Constant + trend	−2.397	−3.362 ***	−1.285	−3.067 **

Notes: Critical Values: For model with constant: −2.60 (1%), −2.34 (5%), −2.21 (10%), For model with constant and trend: −3.15 (1%), −2.88 (5%), −2.74 (10%). ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

.

According to the test results presented in Table 3, although the test statistics for some models (e.g., EC, FD, EG) are lower than the critical values at the level, when considered as a whole, the test statistics in both the constant and the constant-and-trend models exceed the critical values. Therefore, it can be concluded that the variables are not stationary at their levels. As a result, the CIPS test was repeated using the first-differenced series. At the first difference, the test statistics for the variables in both country groups are lower than the critical values at various significance levels, indicating that the differenced series are stationary and do not contain a unit root.

After conducting individual cross-sectional dependence and unit root tests for the variables within each country group, it is also necessary to test the models themselves for cross-sectional dependence and parameter homogeneity. This step ensures the selection of the most appropriate estimation method for subsequent analyses. Equation (1) was specified separately for the E7 and G7 country groups. The results of the cross-sectional dependence and homogeneity tests related to these models are presented in

Table 4. Cross-sectional dependence and homogeneity test results for the model.

	E7 Countries		G7 Countries
Horizontal Cross-sectional Dependence Test	Stat.	Prob.	Stat.	Prob.
LMadj	1.905	0.028 **	7.897	0.000 ***
Homogeneity Tests	Stat.	Prob.	Stat.	Prob.
$\widetilde{∆}$	8.538	0.000 ***	7.301	0.000 ***
${\widetilde{∆}}_{\mathit{a}\mathit{d}\mathit{j}}$	9.931	0.000 ***	8.492	0.000 ***

Note: ***, ** Significant at 1% and 5%, respectively.

.

When examining the results presented in Table 4, it can be seen that the probability value associated with the cross-sectional dependence test for E7 countries is significant at the 5% level, while the probability values for the other tests are significant at the 1% level, leading to the rejection of the null hypotheses. Accordingly, it can be inferred that the E7 and G7 countries exhibit cross-sectional dependence and a heterogeneous structure.

Summarizing the pre-test results used to determine the appropriate methodology for analyzing the relationship among the variables in E7 and G7 countries, it is evident that all variables exhibit cross-sectional dependence, are stationary at first difference, and that the models demonstrate cross-sectional dependence and heterogeneity. In this context, the LM bootstrap cointegration test, which accounts for these conditions, is deemed appropriate. The results of this test are presented in

Table 5. Results of the LM bootstrap cointegration test.

	E7 Countries		G7 Countries
	Constant	Constant + Trend	Constant	Constant + Trend
LM Stat.	3.418	9.809	6.635	11.594
Bootstrap p	0.999	0.100	0.826	0.621
Asymp p	0.000	0.000	0.000	0.000

.

The LM bootstrap cointegration test can be applied in the presence or absence of cross-sectional dependence and provides separate probability values for both scenarios. In cases of cross-sectional dependence, the bootstrap probability value is taken into consideration. As shown in Table 5, the bootstrap probability values are not significant for either group of countries. Therefore, the null hypothesis is accepted, indicating the presence of a cointegration relationship among the variables examined in the analysis for both the E7 and G7 countries. This implies that, in both groups, the variables CO₂ emissions, FD, EG, EC, and TO move together in the long run. To estimate the long-run coefficients among cointegrated variables, a cointegration estimator is employed. In this study, the panel AMG estimator, which accounts for cross-sectional dependence and heterogeneity, was used. The results at the group level are presented in

Table 6. Results of the panel AMG estimator.

	E7 Countries		G7 Countries
	Coefficient	p	Coefficient	p
FD	−0.086	0.262	0.061	0.047 **
EG	0.125	0.078 *	0.413	0.057 *
EC	0.985	0.000 ***	0.699	0.014 ***
TO	−0.024	0.512	−0.079	0.056 *

Note: ***, **, * Significant at 1%, 5%, and 10%, respectively.

.

As shown in Table 6, the probability value for the EG variable is significant at the 10% level, and that of the EC variable is significant at the 1% level in E7 countries, whereas the probability values for FD and TO are not significant. Therefore, a one-unit increase in EG leads to an increase of 0.125 units in CO₂ emissions, and a one-unit increase in EC results in an increase of 0.985 units in CO₂ emissions in the E7 countries. On the other hand, in the G7 countries, the FD variable is significant at the 5% level, the EC variable at the 1% level, and the EG and TO variables at the 10% level. Accordingly, a one-unit change in FD increases CO₂ emissions by 0.061 units, in EG by 0.413 units, and in EC by 0.699 units, while a one-unit change in TO reduces CO₂ emissions by 0.079 units. Based on the panel AMG results, it can be concluded that EG and EC in E7 countries, and FD, EG, and EC in G7 countries, contribute to environmental degradation, whereas TO mitigates it.

The findings indicating that economic growth increases environmental degradation in E7 countries are consistent with the results of studies carried out by Kartal et al. (2025), Gyamfi et al. (2024), U. Khan et al. (2023), Huang et al. (2022), K. Li et al. (2022), Aydoğan and Vardar (2019), and Doğan and Değer (2018). Similarly, the evidence suggesting that energy consumption exacerbates environmental degradation in E7 countries corroborates the findings reported by Liang et al. (2024), Husnain et al. (2022), K. Li et al. (2022), Aydoğan and Vardar (2019), Doğan and Değer (2018), and Doğan et al. (2022).

For G7 countries, the results showing that economic growth accelerates environmental degradation are consistent with the conclusions emphasized by H. Chen et al. (2025), Elhassan (2025), Gyamfi et al. (2024), U. Khan et al. (2023), Kostakis and Arauzo-Carod (2023), Zaman and Yu (2023), and Murshed et al. (2022). The evidence that energy consumption accelerates environmental degradation in G7 economies aligns with Elhassan (2025), while the findings that financial development intensifies environmental degradation are supported by Xu and Xu (2024), and Huang et al. (2022).

The AMG estimator provides long-term coefficient estimates both at the panel and country levels. The country-level results are presented in

Table 7. Results of the AMG estimator at the country level.

	Brazil		Canada		China		United Kingdom		USA
	Coef.	p	Coef.	p	Coef.	p	Coef.	p	Coef.	p
FD	−0.348	0.005 ***	−0.054	0.140	−0.178	0.013 **	0.163	0.002 ***	−0.021	0.832
EG	0.134	0.595	0.230	0.372	−0.076	0.727	0.088	0.401	−0.012	0.918
EC	1.414	0.000 ***	0.640	0.016 **	1.169	0.000 ***	0.577	0.000 ***	1.261	0.000 ***
TO	0.093	0.060 *	0.101	0.084 *	−0.080	0.500	−0.064	0.330	−0.004	0.890
	Indonesia		France		India		Japan		Germany
	Coef.	p	Coef.	p	Coef.	p	Coef.	p	Coef.	p
FD	−0.238	0.182	0.045	0.523	−0.073	0.215	0.051	0.531	0.096	0.285
EG	0.516	0.010 **	−0.224	0.469	0.002	0.991	1.475	0.000 ***	0.651	0.002 ***
EC	0.774	0.000 ***	1.387	0.000 ***	1.122	0.000 ***	−0.863	0.000 ***	0.953	0.000 ***
TO	−0.037	0.642	−0.165	0.129	−0.048	0.012 **	−0.047	0.343	−0.213	0.018 **
	Mexico		Italy		Russia		Türkiye
	Coef.	p	Coef.	p	Coef.	p	Coef.	p
FD	−0.082	0.582	0.147	0.129	0.040	0.105	0.276	0.126
EG	0.107	0.594	0.686	0.000 ***	0.095	0.038 **	0.095	0.656
EC	0.846	0.000 ***	0.938	0.000 ***	0.885	0.000 ***	0.683	0.000 ***
TO	−0.138	0.048 **	−0.158	0.002 ***	0.121	0.013	−0.076	0.465

Note: ***, **, * Significant at 1%, 5%, and 10%, respectively.

.

The test results presented in Table 6 can be interpreted on a country-by-country basis as follows:

Brazil: The probability values of FD and EC are significant at the 1% level, while that of TO is significant at the 10% level. The probability value of EG, however, is not significant. Accordingly, a one-unit change in FD reduces CO₂ emissions by 0.348 units, while a one-unit change in EC increases CO₂ emissions by 1.414 units, and a one-unit change in TO increases CO₂ emissions by 0.093 units. EG appears to have no significant effect on CO₂ emissions.

China: The probability value of FD is significant at the 5% level, and that of EC is significant at the 1% level. The other variables are not significant. Based on these findings, a one-unit change in FD reduces CO₂ emissions by 0.178 units, while a one-unit change in EC increases CO₂ emissions by 1.169 units. The other variables do not have a significant impact on CO₂ emissions.

Indonesia: EG is significant at the 5% level, and EC at the 1% level, whereas the other variables are not significant. A one-unit change in EG increases CO₂ emissions by 0.516 units, and a one-unit change in EC increases CO₂ emissions by 0.774 units. The remaining variables do not influence CO₂ emissions.

India: EC is significant at the 1% level, and TO at the 5% level. The probability values of the other variables are not significant. Accordingly, a one-unit change in EC increases CO₂ emissions by 1.122 units, while a one-unit change in TO reduces CO₂ emissions by 0.048 units. The remaining variables have no significant effect on CO₂ emissions.

Mexico: EC is significant at the 1% level and TO at the 5% level, whereas the other variables are not significant. Based on these results, a one-unit change in EC increases CO₂ emissions by 0.846 units, while a one-unit change in TO reduces CO₂ emissions by 0.138 units. The other variables do not significantly affect CO₂ emissions.

Russia: EG is significant at the 5% level and EC at the 1% level. The remaining variables are not significant. A one-unit change in EG increases CO₂ emissions by 0.095 units, and a one-unit change in EC increases CO₂ emissions by 0.885 units. The other variables do not have a significant effect.

Türkiye: Only EC is significant at the 1% level. The other variables are not significant. This indicates that a one-unit change in EC increases CO₂ emissions by 0.683 units, while the other variables do not have a significant effect on CO₂ emissions.

USA: EC is significant at the 1% level, while the other variables are not significant. These findings suggest that a one-unit change in EC increases CO₂ emissions by 1.261 units, whereas the other variables have no significant effect.

Germany: EG and EC are significant at the 1% level, and TO is significant at the 5% level. FD, however, is not significant. Accordingly, a one-unit change in EG increases CO₂ emissions by 0.651 units, EC by 0.953 units, and a one-unit change in TO reduces CO₂ emissions by 0.213 units. FD does not appear to affect CO₂ emissions.

France: EC is significant at the 1% level. The remaining variables are not significant. Thus, a one-unit change in EC increases CO₂ emissions by 1.387 units, while the other variables do not significantly influence emissions.

United Kingdom: Both FD and EC are significant at the 1% level, while the other variables are not. A one-unit change in FD increases CO₂ emissions by 0.163 units, and a one-unit change in EC increases emissions by 0.577 units. The other variables do not affect CO₂ emissions.

Italy: EG, EC, and TO are significant at the 1% level, whereas FD is not. A one-unit change in EG increases CO₂ emissions by 0.686 units, a one-unit change in EC increases emissions by 0.938 units, and a one-unit change in TO reduces emissions by 0.158 units. FD has no significant effect.

Japan: EG and EC are significant at the 1% level, while the other variables are not significant. A one-unit change in EG increases CO₂ emissions by 1.475 units, whereas a one-unit change in EC reduces emissions by 0.863 units.

Canada: EC is significant at the 5% level and TO at the 10% level. The other variables are not significant. The findings indicate that a one-unit change in EC increases CO₂ emissions by 0.640 units and a one-unit change in TO increases emissions by 0.101 units. The other variables do not significantly affect CO₂ emissions.

In Brazil and China, financial development contributes to reducing CO₂ emissions. This result may be attributed to capitalization, technological, income, and regulatory effects. In this regard, firms may channel low-cost financial resources into environmentally friendly projects, adopt less carbon-intensive technologies, respond to consumers’ growing preference for green products and services, and benefit from increased bank lending directed toward financing environmentally sensitive projects. Conversely, in the United Kingdom, financial development may have increased CO₂ emissions through capitalization, technological, and income effects. In this case, easily accessible low-cost funds in a mature financial market may have stimulated production, modern technological advances may have introduced new polluting elements, and rising household income may have led to greater demand for energy-intensive goods, thereby exacerbating environmental degradation.

In Indonesia, Japan, Germany, Italy, and Russia, economic growth was shown to increase CO₂ emissions. This can be explained by the intensified production and industrialization efforts aimed at sustaining economic growth, which in turn has contributed to greater environmental pollution. Moreover, green economy initiatives seem not to have been sufficiently widespread in these countries.

While energy consumption accelerates environmental degradation in countries other than Japan, in Japan it appears to mitigate such degradation. This may be due to Japan’s increasing shift from fossil fuels to renewable energy sources. In particular, recent statements by the Japanese government regarding the implementation of energy policies targeting carbon neutrality by 2050 through emission reductions in the electricity, industrial, and transportation sectors (U.S. Energy Information Administration, 2024) further support these findings. In Brazil, China, Indonesia, India, Mexico, Russia, and Turkey, as well as in Canada, Germany, France, the United Kingdom, Italy, and the USA, fossil energy sources are used more extensively than renewable ones, thereby contributing to rising CO₂ emissions.

Trade openness was found to increase CO₂ emissions in Brazil and Canada, which may be explained by scale and composition effects. The expansion of international trade in these countries may have amplified production, thereby aggravating environmental degradation. In addition, the importation of pollution-intensive, low-cost technologies and the reliance on pollution-heavy production processes may have further deteriorated environmental quality. By contrast, trade openness seems to reduce CO₂ emissions through technological effects in India, Germany, Mexico, and Italy. This suggests that these countries have been importing and employing cleaner technologies, which has helped mitigate environmental degradation.

Panel and country-level causality analyses were conducted following the cointegration tests, and the results are presented in

Table 8. Causality analysis results for FD, EG and CO2.

	FD→CO2		CO2→FD		EG→CO2		CO2→EG
E7 Countries	Wald	p	Wald	p	Wald	p	Wald	p
Brazil	5.670	0.190	2.464	0.250	0.055	0.900	11.223	0.030 **
China	1.497	0.210	9.273	0.360	3.136	0.150	8.189	0.040 **
Indonesia	1.856	0.060 *	9.961	0.060 *	0.011	0.950	1.604	0.230
India	4.662	0.040 **	1.859	0.220	0.118	0.980	0.056	0.840
Mexico	3.303	0.080 *	4.822	0.250	16.518	0.090 *	0.258	0.870
Russia	0.092	0.700	0.058	0.750	1.764	0.420	14.388	0.010 ***
Türkiye	1.426	0.980	5.631	0.960	1.383	0.760	0.320	0.700
Panel	24.313	0.042 **	16.901	0.262	11.248	0.666	26.941	0.020 **
G7 Countries	Wald	p	Wald	p	Wald	p	Wald	p
Canada	0.050	0.900	0.112	0.740	4.082	0.280	14.413	0.080
Germany	6.569	0.130	1.046	0.890	0.859	0.870	0.065	0.770
France	0.470	0.710	1.583	0.460	4.494	0.410	0.054	0.780
United Kingdom	1.997	0.710	1.130	0.900	2.492	0.520	1.392	0.480
Italy	0.343	0.710	7.531	0.140	0.085	0.980	1.657	0.350
Japan	3.696	0.320	0.151	0.890	0.098	0.980	2.228	0.480
USA	1.623	0.360	2.498	0.150	0.214	0.980	4.933	0.110
Panel	10.668	0.712	10.559	0.720	6.037	0.966	15.521	0.343

Note: ***, **, * Significant at 1%, 5%, and 10%, respectively.

and

Table 9. Causality analysis results for EC, TO and CO2.

	EC→CO2		CO2→EC		TO→CO2		CO2→TO
E7 Countries	Wald	p	Wald	p	Wald	p	Wald	p
Brazil	2.175	1.000	5.688	0.910	3.133	0.270	0.002	0.930
China	3.342	1.000	1.432	0.990	0.788	0.550	14.965	0.000 ***
Indonesia	4.175	0.990	0.489	1.000	4.904	0.740	9.940	0.650
India	3.164	1.000	0.720	1.000	3.758	0.410	2.197	0.020 **
Mexico	2.244	0.930	0.725	1.000	4.670	0.040 **	0.146	0.830
Russia	9.365	0.950	0.004	1.000	10.976	0.070 *	14.681	0.080 *
Türkiye	8.194	0.990	0.026	1.000	0.019	0.920	8.965	0.360
Panel	0.288	1.000	0.209	1.000	18.123	0.201	20.568	0.000 ***
G7 Countries	Wald	p	Wald	p	Wald	p	Wald	p
Canada	0.087	0.920	2.832	0.640	0.016	0.930	0.573	0.770
Germany	43.730	0.420	17.403	0.480	1.609	0.250	16.678	0.000 ***
France	1.179	1.000	27.631	0.900	6.394	0.240	14.602	0.010 ***
United Kingdom	1.637	1.000	5.549	1.000	4.419	0.190	2.891	0.860
Italy	16.041	0.970	13.163	0.990	15.534	0.000 ***	27.694	0.000 ***
Japan	0.191	0.980	0.136	1.000	6.149	0.030 **	0.280	0.630
USA	26.447	0.490	0.066	0.960	8.815	0.010 ***	0.008	0.980
Panel	3.430	0.998	2.673	1.000	40.652	0.000 ***	41.320	0.000 ***

Note: ***, **, * Significant at 1%, 5%, and 10%, respectively.

.

As shown in Table 8, bidirectional causality between FD and CO₂ emissions is observed in Indonesia. A unidirectional causality running from FD to CO₂ emissions is found for India, Mexico, and the overall panel of E7 countries. Moreover, a unidirectional causality from EC to CO₂ emissions is found in Mexico, whereas a unidirectional causality from CO₂ emissions to EC is found for Brazil, China, Russia, and the E7 panel as a whole. In contrast, among G7 countries, only Canada shows a unidirectional causal relationship from CO₂ emissions to EC. No significant causality is detected between FD or EC and CO₂ emissions in the remaining G7 countries.

Given the results in Table 9, no causality is detected between EC and CO₂ emissions in the E7 countries. However, a bidirectional causal relationship is observed between TO and CO₂ emissions in Russia, while a unidirectional causality from TO to CO₂ emissions is evident in Mexico. In China, India, Russia, and the overall E7 panel, unidirectional causality is found running from CO₂ emissions to TO. For G7 countries, there is no evidence of causality between EC and CO₂ emissions. Nonetheless, a bidirectional causality between TO and CO₂ emissions is found for both Italy and the G7 panel. Furthermore, a unidirectional causality from TO to CO₂ emissions is observed in Japan and the USA, while Germany and France exhibit unidirectional causality from CO₂ emissions to TO.

At the panel level, evidence of a unidirectional causality running from CO₂ emissions to economic growth in China, Brazil, and Russia is consistent with the results reported by Atılgan and İspir (2021), Tong et al. (2020), and Aydoğan and Vardar (2019).

6. Conclusions

The increasing demand for natural resources has placed substantial pressure on ecosystems, resulting in major environmental issues such as climate change, land degradation, water and air pollution, loss of biodiversity, and global warming (Chu & Lee, 2022). The problem of environmental degradation has become increasingly significant for developing, emerging, and developed economies. The distinct nature of the environmental threats faced by different countries, combined with the irresistible drive toward achieving higher EG, severely hinders genuine efforts to control EC and carbon emissions (Pradhan et al., 2024). This situation necessitates the examination of environmental degradation across country groups with varying development levels and the formulation of appropriate policy recommendations. In this context, the present study investigates the relationship between FD, EG, EC, TO, and environmental degradation in G7 and E7 countries for the period 2000–2021 using panel cointegration and causality analyses.

This study utilizes a panel dataset composed of 22 years of data from both country groups. To ensure the validity of the panel data analyses, preliminary tests for cross-sectional dependence and unit roots were conducted. The results indicate that all variables exhibit cross-sectional dependence and are stationary at the first difference in both groups. Additionally, evidence was found that the established models also exhibit cross-sectional dependence and that the panels have a heterogeneous structure.

After the preliminary tests, the LM Bootstrap cointegration test, which accounts for cross-sectional dependence, was implemented. The findings suggest that the variables in both country groups move together in the long run. The long-run coefficients of the identified cointegration relationship were estimated using the panel AMG estimator. At the panel level, the AMG results revealed that EG and EC accelerate environmental degradation in E7 countries, whereas in G7 countries, FD, EG, and EC all contribute to environmental degradation, while TO mitigates it. Finally, the country-level panel AMG test results revealed that financial development mitigates environmental degradation in Brazil and China, while it exacerbates it in the United Kingdom. Economic growth accelerates environmental degradation in Indonesia, Japan, Germany, Italy, and Russia. Energy consumption reduces environmental degradation in Japan but intensifies it in all other E7 countries. Trade openness slows environmental degradation in Italy, Germany, India, and Mexico, whereas it aggravates it in Canada and Brazil.

Panel causality analysis results suggest that, in E7 countries, changes in FD drive CO₂ emissions, while changes in CO₂ emissions affect EG and TO. At the country level, mutual causality was found between FD and CO₂ emissions in Indonesia and between TO and CO₂ emissions in Russia. Moreover, the findings indicate that FD influences CO₂ emissions in India, FD, EG, and TO all impact CO₂ emissions in Mexico. CO₂ emissions, in turn, affect EG in Brazil, China, and Russia, and influence TO in China and India. No causality was found between EC and CO₂ emissions. In G7 countries, mutual causality between TO and CO₂ emissions was observed at the panel level. In addition, bidirectional causality was also found between TO and CO₂ emissions in Italy. The results further suggest that changes in TO affect CO₂ emissions in Japan and the USA; changes in CO₂ emissions affect EG in Canada, and TO in Germany and France. No evidence was found for a causal relationship between FD, EG, or EC and CO₂ emissions in the remaining cases.

Within the scope of this study, the findings regarding the relationship between FD, EG, EC, TO, and environmental degradation suggest that there are both similarities and differences between E7 and G7 countries. In terms of cointegration results, both country groups exhibit a long-term relationship among the variables, with EG and EC contributing to increased environmental degradation. Similarly, the causality analyses reveal no causal relationship between EC and environmental degradation, while a unidirectional causality is observed from CO₂ emissions to TO. However, in contrast to the E7 countries, the G7 countries demonstrate some distinct characteristics. In the long run, FD appears to exacerbate environmental degradation, while TO tends to mitigate it. Furthermore, no causal relationship is found between FD or EG and CO₂ emissions in the G7 group, constituting the key differences revealed by the analysis.

Given the results achieved, it can be stated that in both E7 and G7 countries, fossil fuel-based energy sources, which are harmful to the environment, are heavily utilized in the execution of economic activities, leading to environmental damage. From this perspective, it is advisable for these country groups to increase the use of environmentally friendly renewable energy sources in their economic operations. In this context, recommended policy measures include implementing regulations that promote clean energy sources, raising awareness among firms and individuals, developing financial instruments to support the financing of renewable energy projects, and formulating green economy strategies that foster EG without compromising environmental sustainability.

Additionally, the findings confirm that in G7 countries, the financial sector often provides funding for environmentally harmful projects and activities, while green financial instruments that support the environment are not widely adopted. Nevertheless, it is also observed that in the context of international trade, companies in G7 countries help reduce environmental degradation by employing environmentally conscious clean technologies and renewable energy sources. Therefore, in order to improve environmental quality in G7 countries, it is recommended to develop green finance mechanisms within the financial sector, to encourage firms and individuals to use environmentally friendly financial instruments, and to financially support environmentally sustainable projects and technologies.

Although this study presents significant findings and policy recommendations for both E7 and G7 countries, it is subject to certain limitations. The use of only CO₂ emissions as the indicator of environmental degradation, the limited number of variables affecting the environment, and the reliance on standard methodologies over a fixed time period may constrain the generalizability of the findings. Moreover, the exclusion of various dynamic factors that could influence the investigated relationships might have affected the results.

In light of these limitations, future studies are advised to incorporate a broader range of environmental indicators, to include additional variables such as urbanization, population growth, and foreign direct investment, to analyze longer time periods and more extensive samples of developed and developing country groups, and to apply a wider variety of methodological approaches to generate more detailed and robust evidence.