The Relevance of Financial Development, Natural Resources, Technological Innovation, and Human Development for Carbon and Ecological Footprints: Fresh Evidence of the Resource Curse Hypothesis in G-10 Countries

Abstract

This study focuses on the effect of financial development, natural resource rent, human development, and technological innovation on the ecological and carbon footprints of the G-10 countries between 1990 and 2022. This study also considers the impact of globalization, trade openness, urbanization, and renewable energy on environmental degradation. The study uses Kao and Westerlund DH cointegration tests, FMOLS and DOLS estimators, and panel Fisher and Hatemi-J asymmetric causality tests to provide reliable results. Long-run estimates confirm an inverted U-shaped linkage between financial development and ecological and carbon footprints. Natural resource rent and technological innovation increase ecological and carbon footprints, while human development decreases them. Furthermore, globalization, trade openness, and renewable energy contribute to environmental quality, while urbanization increases environmental degradation. The Fisher test findings reveal that financial development, natural resource rent, human development, and technological innovation have a causal link with the ecological and carbon footprint. The results of the Hatemi-J test show that the negative shocks observed in the ecological and carbon footprint are affected by both negative and positive shocks in financial development, natural resource rent, and technological innovation. Moreover, positive and negative shocks in human development are the main drivers of negative shocks in the carbon footprint, while positive shocks in human development lead to negative shocks in the ecological footprint.

1. Introduction

Governments worldwide have established carbon neutrality goals as a top priority to address the adverse effects of climate change and promote sustainable environmental development, all while ensuring energy security. However, balancing environmental conservation and economic growth is complex for developed and developing countries. Economic, social, and environmental activities have led to a trade-off between economic performance and environmental sustainability [1,2,3]. Thus, the connection between environmental issues and economic development is essential in the existing literature. Financial development has emerged as a significant factor related to environmental changes because of the growing role of financial markets in economies [4,5]. It positively impacts production capacities by enabling investments and creating mechanisms for sustainable economic growth. Still, FD can also adversely affect the environment by increasing natural resource consumption and promoting credit use for purchasing machinery, electrical appliances, cars, and real estate [5,6,7]. Countries with advanced financial systems may contribute to environmental pollution through unsustainable practices like the overuse of natural resources [8]. By making credit more accessible, financial systems can encourage businesses to purchase high-emission and energy-consuming products that are costly. On the other hand, stock market development can decrease financial costs and enhance firms’ liquidity and efficiency, leading to increased energy consumption and more profound environmental impacts [9].

Many studies indicate that financial development could lead to environmental degradation [3,10,11,12]. Yet, other research suggests that a positive linkage exists between financial development and environmental quality [13,14,15,16]. In line with this idea, financial development enhances environmental quality by attracting eco-friendly projects through research and development support while encouraging investments in clean technologies [6]. Some recent studies [17,18] have found that financial development has the scale and structural changes that affect the environment. Their findings showed that early stages of financial development result in increased environmental degradation due to scale effects. However, after reaching a certain threshold, financial development influences transformation, leading to improved environments via efficient natural resource usage. Consequently, the link between environmental pollution and financial development remains uncertain and shifts depending on how developed countries are.

Conversely, economies dependent on natural resources often have weak demands for governance and political reforms, as they prioritize resource extraction over human capital investments. This dependence can negatively affect the economic structure by hampering industrialization and modernization. The resource curse is associated with the emergence of problems such as corruption and weak governance in natural resource-rich countries, as opposed to the expected economic prosperity. The natural resource curse can manifest in different ways in developed and developing nations, hindering economic growth. Various factors, such as unfavorable exchange rate policies, weak institutional structure, insufficient capital accumulation, domestic political instability, and conflicts, cause this phenomenon. Solving these fundamental problems is crucial to maximize financial gains across countries [19,20]. On the other hand, the unsustainable use of natural resources reduces a country’s biocapacity and expands its ecological footprint, eventually leading to an ecological deficit. Accelerating economic growth can negatively impact ecosystems by encouraging the more intensive extraction of natural resources. However, the transition from traditional to advanced technologies has the potential to improve environmental quality while supporting economic development. Modern technologies can mitigate these negative impacts by providing alternative solutions to natural resources through value-added products, artificial resources, recycling, reprocessing, and innovative processes [21].

Nowadays, emphasis is placed on how pursuing skilled human capital and education facilitates sustainable, eco-friendly advancements and supports green technological readiness [22]. Human advancements exhibit various dimensions that positively impact the environment; for example, obtaining skilled labor increases productivity while promoting new technology integrations. This ultimately leads to enhanced productivity, which automatically feeds back into the development of education that has the necessary knowledge of current ecological concerns, and this increases the protective awareness of the environment. However, current reviews show that the harmful effects of human actions continue to be the primary cause of pollution problems occurring in ecosystems [12,23].

Based on the above discussion, this study examines the effect of financial development, natural resource rent, technological innovation, and human development on ecological and carbon footprints, particularly in the G-10, a group of industrialized countries. The G-10 countries are a group of eleven industrialized countries with similar economic interests. The main objective of this group is to coordinate monetary and fiscal policies to promote global economic stability. The countries in the group have experienced strong economic growth. However, this rapid economic growth can lead to an increased consumption of natural resources, with negative impacts on environmental quality. Recognizing this, G-10 countries have recently increased their investments in renewable energy projects and environmentally friendly technologies [2,24]. In this context, this study specifically aims to address the following research questions: (1) Is there a non-linear relationship between financial development and environmental sustainability? (2) Does the abundance of natural resources negatively affect environmental sustainability, supporting the resource curse hypothesis? (3) Do human development and technological innovation play an effective role in ensuring environmental sustainability?

This study contributes to the existing literature in four ways: (1) It provides new evidence by investigating whether there is an inverted U-shaped connection between financial development and environmental degradation. G-10 countries represent financially developed countries with strong capital markets and banking systems. It is therefore important to assess whether such investments offset the negative impact of FD on environmental degradation. (2) It analyzes the connection between natural resource rents and environmental degradation in industrialized countries and evaluates it within the framework of the resource curse hypothesis in developed economies. The natural resource curse hypothesis argues that economies overly dependent on natural resource rents may ignore industrialization and economic diversification and relegate environmental protection policies to the background [19]. In this context, it is important to assess the impact of natural resource wealth on efforts to prevent environmental degradation. (3) This study has broad policy implications for environmental sustainability and development by including human development and technological innovation in the model. Moreover, this study takes a multifaceted view of environmental sustainability and uses ecological and carbon footprint indicators of environmental pollution. (4) This study employs FMOLS and DOLS to obtain reliable and appropriate regression analysis results. Additionally, the findings are supported by symmetric and asymmetric causality tests. This study offers broad explorations for government policy elaboration for sustainable growth in this context.

The following section provides an overview of the existing literature, literature gap, and hypothesis development. Section 3 comprehensively discusses the variables, model specification, and methodological approach. Section 4 contains the empirical findings, while Section 5 focuses on developing policy recommendations.

2. Literature Review

While a strong financial sector is a key driver of economic growth, its environmental impacts must be assessed. Rapid industrialization and traditional energy use can harm the environment, as natural resources are depleted for various reasons [25,26]. However, a strong financial system can facilitate capital growth and technological advances by converting savings into effective investments [27]. These conditions help stimulate investments in research and development. This focus can improve environmental quality by promoting clean energy technologies [7,11,17]. However, analyzing this relationship can be complex, as various factors influence energy consumption. The lack of financial resources and the level of public awareness of energy issues play an important role in shaping energy consumption. However, the operational processes of the finance department can also create environmental impacts. However, promoting digitalization and using energy-efficient technologies can play an important role in reducing the environmental impact of the finance department and contribute to adopting sustainable financial management [28,29]. There is evidence for both views in existing studies. Across different periods and samples, evidence suggests that environmental quality can be improved through financial development and vice versa [3,10,13,30]. Moreover, from a new perspective, some studies show an inverted U-shaped connection between financial development and environmental pollution indicators, as shown in Table 1 [17,18,31].

When countries mainly focus on economic growth, they often use natural resources in an unsustainable way. Unsustainable resource management leads to serious environmental problems. In simpler terms, using natural resources stimulates growth but is closely linked to the healthy development of economies. In the early stages of development, resource extraction tends to be high without much concern about environmental impacts. Therefore, as production accelerates with the burning of fossil fuels, there are negative impacts on ecology [15,32,33]. On the other hand, although natural resource rent has a crucial place in the country’s economy, inequality in its distribution increases the risk of corruption by encouraging favoritism. This makes it challenging to allocate natural resources transparently and efficiently, negatively affecting environmental sustainability. Resource scarcity exacerbates environmental degradation as human demand increases, raising extraction costs and increasing natural resource rents. Moreover, inequitable resource allocation and consumption in excess of what nature produces threaten sustainability by creating an environmental deficit [34,35]. In contrast, some studies suggest that natural resource rents can reduce environmental degradation. For example, natural resources can support globalization by expanding foreign trade volume, encouraging foreign investment, and increasing foreign exchange flows. It can also reduce carbon emissions through greater efficiency in energy use. Moreover, the abundance of natural resources can limit carbon emissions by contributing to a reduction in fossil fuel consumption and imports. According to this view, natural resource rent is considered a mechanism to secure the sustainable development of ecological resources [36]. However, there is no consensus on this in the literature, and studies provide different evidence, often based on timeline variations and sample contexts, as expressed in Table 1.

Technological innovation improves energy efficiency, reducing consumption and mitigating environmental impacts. Moreover, innovative technologies can reduce dependence on fossil fuels by stimulating the development of renewable energy sources. However, the effectiveness of this process in reducing the ecological footprint depends on the institutional quality, including energy policies and regulations [37]. However, another view argues that the use of technology in economic growth and the expansion of production activities that negatively impact environmental sustainability can increase environmental degradation [38]. More specifically, if technological progress is associated with environmental damage, the rebound effect can favor and disadvantage these advances [39]. The rebound effect is defined as the economic gains from increased resource efficiency through innovation, but with a consequent increase in resource consumption. Increased efficiency in resource use can lead to lower prices and, hence, higher demand. This effect can occur directly through price elasticity or indirectly through a shift in purchasing power towards products and services that use natural resources [40]. In the context of these discussions, the empirical findings on the effect of technological innovation on environmental degradation in the existing literature (Table 1) are mixed.

In addition to economic factors, human development influences natural resource usage, which pertains directly to environmental sustainability. Specifically, pressures on key human development components, wealth, health, education, etc., often impact environmental quality negatively due to unsustainable consumption and production patterns, globalization, and population growth [41]. However, educational initiatives can foster positive environmental outcomes, promoting sustainability [42]. Human well-being and environmental improvement are intricately linked. The effective management of environmental challenges can lead to better human outcomes. Ensuring that human well-being improves while reducing pressure on the environment is a crucial aim of sustainable development [43]. Liu et al. [42] assert that educating people can enhance environmental quality and foster sustainable outcomes. It underscores the link between human well-being, development, and environmental improvement. Managing environmental issues effectively contributes to better human outcomes. Therefore, improving human well-being while halting natural resource pressures is key to sustainable development goals [43]. In summary, attaining high human development with low ecological footprints epitomizes ideal sustainable development [44]. Existing empirical evidence typically supports this perspective (Table 1).

Conversely, the extant literature has associated globalization, trade openness, urbanization, and renewable energy with environmental degradation. Consequently, these factors, which are believed to affect carbon and ecological footprints, are considered control variables in the study.

Globalization is about how countries connect with the global economy. This connection helps balance supply and demand in markets. Because globalization directly influences production processes in economies, it also affects environmental quality [45]. The trade volume driven by globalization has various impacts on the environment. These come from income, technology, and composition effects [25,46]. The income effect means that globalization can lead to higher carbon emissions, which decreases environmental quality [46]. Conversely, the technical effect suggests that globalization might positively affect the environment by promoting eco-friendly technologies [33]. Lastly, a composition effect shows how globalization changes an economy’s capital and labor mix, impacting its structure [46]. Many studies have shown varying results about globalization and environmental quality [27,47,48].

The accessibility of foreign direct investment and international trade has increased due to globalization. However, whether trade affects environmental quality positively or negatively depends on multiple aspects, such as development level, industry type, and the extent of globalization within an economy. Two key factors, scale and composition, explain this impact. The scale effect indicates that as an economy grows through trade, pollution can rise due to greater resource consumption. Conversely, the technical effect points to better environmental quality as incomes rise, primarily when focusing on clean sectors in developed nations. However, countries prioritize growth initially, even at their environment’s expense, when they are just beginning to develop [49,50,51]. Empirical data support the view that trade openness can lead to positive [51,52,53] and negative impacts on environmental quality across different settings [42,54,55].

Migration from rural to urban areas has sparked urbanization, a socio-economic transformation process concentrated on better job opportunities, health, and education [56]. Urbanization triggers modernization, fueling economic and social growth [32,57]. The impacts of urbanization on environmental quality are complex. They arise from infrastructure construction and factors like population, housing, industrialization, transport demand, and trade intensity [58]. Urbanization can worsen environmental degradation by escalating fossil fuel consumption through higher industrial demands. Conversely, it might reduce environmental harm by promoting renewable energy due to the increased purchasing power of urban folks [59].

Conversely, the negative impacts of fossil fuel consumption on the environment and welfare reveal the necessity of turning to alternative energy sources without disrupting economic growth. Renewable energy stands out as an alternative that supports sustainable development and should be promoted. With a lower ecological footprint than fossil fuels, renewable energy sources are expected to play a major role in ensuring environmental sustainability. In this context, many countries have recognized the importance of renewable energy and have developed various policies to increase its use. However, while renewable energy sources have the potential to improve environmental quality, some countries are struggling to fully realize this transformation due to potential impacts on economic growth and socio-economic barriers. While it is generally accepted that renewable energy effectively reduces carbon emissions, its effectiveness in promoting the decoupling of the ecological footprint and economic growth has yet to be established [51,60,61,62]. Nevertheless, empirical evidence generally confirms that renewable energy reduces environmental degradation [37,63,64].

Table 2 shows some empirical research highlighting the link between globalization, trade openness, urbanization, renewable energy, and ecological and carbon footprints.

Table 1. Summary of the literature.

Financial Development—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Godil et al. [30]	Turkiye	1986–2018	QARDL	FD ↑ EF
Omoke et al. [27]	Nigeria	1971–2014	NARDL	FD ↓ EF
Usman and Hammar [65]	APEC	1990–2017	FGLS, AMG, CCEMG	FD ↓ EF
Usman and Makhdum [10]	BRICS-T	1990–2018	MG, AMG, CCEMG	FD ↑ EF
Yao et al. [13]	BRICS and Next-11	1995–2014	GMM	FD ↓ EF
Saqib [66]	63 Emerged and Developed Economies	1990–2020	AMG, CCEMG	FD ↓ CF
Ashraf et al. [17]	124 Economies	1993–2013	GMM	FD ∩ EF
Khan et al. [31]	APEC	1990–2016	CCEMG	FD ∩ EF
Jahanger et al. [15]	73 Developing Countries	1990–2016	PMG, ARDL	FD ↓ EF
Alam et al. [67]	Oman	1984Q1–2018Q4	ARDL	Short run FD ↑ CF Long run FD ↓ CF
Sun et al. [18]	South Asian	2010–2018	CS-ARDL	FD ∩ CF
Wang et al. [3]	14 Developing European Union	1995–2020	AMG, CCEMG	FD ↑ EF
Naqvi et al. [11]	APEC	1990–2017	DK, FMOLS	FD ↑ EF
Ozturk et al. [68]	South Asian	1971–2018	FMOLS, DOLS	FD ↓ EF
Saqib et al. [69]	Top ten countries with the biggest EF	1990–2109	CCEMG	FD ↑ EF
Yasin et al. [29]	BRICS	1995–2022	DK	FD ↓ EF
Natural Resource Rent—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Danish et al. [70]	BRICS	1992–2016	FMOLS, DOLS	NR ↓ EF
Ahmed et al. [32]	China	1970–2016	ARDL	NR ↑ EF NR ↑ CF
Ulucak et al. [25]	Top 15 Renewable Energy Consumption Economies	1996–2018	PSTR	NR ↑ EF
Ahmad et al. [71]	45 Resource-Rich Countries of Asia	1990–2018	POLS, DK	NR ↓ EF
Ullah et al. [72]	73 Developing Countries	1990–2016	PMG-ARDL	NR ↑ EF
Awosusi et al. [33]	India	1990–2016	ARDL	NR ↓ EF
Onifade [73]	OECD	2000–2019	Panel Quantile	NR ↑ EF
Dao et al. [74]	OECD	2009–2019	MQR	NR ↑ EF
Shittu et al. [75]	BRICS	1992–2018	FMOLS, DOLS, FE-OLS	NR ↑ EF
Human Development—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Kassouri and Altintas [41]	MENA	1990–2016	CCEMG	HD ↓ EF
Pata et al. [1]	Top Ten with Largest EF Economies	1992–2016	AMG	HD ↓ EF
Liu et al. [42]	G-7	1992–2018	CUP-FM, CUP-BC	HD ↓ EF
Qiu and Wan [43]	BRICS	1995–2019	CS-ARDL	HD ↓ EF
Balsalobre-Lorente et al. [76]	G-7	1991–2018	CUP-FM	HD ↓ EF
Nguea and Fotio [77]	31 African Countries	1996–2018	Panel Quantile	HD ↓ EF
Technological Innovation—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Sahoo and Sethi [78]	Newly Industrialized Countries	1990–2017	MG, PMG, AMG	TEC ↓ EF
Chunling et al. [79]	Pakistan	1992–2018	ARDL	TEC ↑ EF
Jahanger et al. [15]	Developing Countries	1990–2016	PMG-ARDL	TEC ↓ EF
Usman and Radulescu, [80]	highest nuclear energy-producing countries	1990–2019	AMG, CCEMG	TEC ↑ CF
Bashir et al. [81]	Newly Industrialized Countries	1990–2018	CS-ARDL	TEC ↓ EF
Dai et al. [82]	ASEAN	1995–2018	CUP-FM, CUP-BC	TEC ↓ EF
Raza et al. [37]	G-20	1990–2021	CS-ARDL	TEC ↓ EF
Chopra et al. [83]	5 high-emitting countries	1990–2022	CS-ARDL	TEC ↓ CF
Quing et al. [63]	South Asian countries	1990–2020	CCEMG	TEC ↓ EF
Nathaniel et al. [84]	Emerging Countries	2000–2020	AMG	TEC ↓ EF
Tiwari et al. [64]	USA	1990–2021	ARDL	TEC ↓ EF

Table 1. Summary of the literature.

Financial Development—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Godil et al. [30]	Turkiye	1986–2018	QARDL	FD ↑ EF
Omoke et al. [27]	Nigeria	1971–2014	NARDL	FD ↓ EF
Usman and Hammar [65]	APEC	1990–2017	FGLS, AMG, CCEMG	FD ↓ EF
Usman and Makhdum [10]	BRICS-T	1990–2018	MG, AMG, CCEMG	FD ↑ EF
Yao et al. [13]	BRICS and Next-11	1995–2014	GMM	FD ↓ EF
Saqib [66]	63 Emerged and Developed Economies	1990–2020	AMG, CCEMG	FD ↓ CF
Ashraf et al. [17]	124 Economies	1993–2013	GMM	FD ∩ EF
Khan et al. [31]	APEC	1990–2016	CCEMG	FD ∩ EF
Jahanger et al. [15]	73 Developing Countries	1990–2016	PMG, ARDL	FD ↓ EF
Alam et al. [67]	Oman	1984Q1–2018Q4	ARDL	Short run FD ↑ CF Long run FD ↓ CF
Sun et al. [18]	South Asian	2010–2018	CS-ARDL	FD ∩ CF
Wang et al. [3]	14 Developing European Union	1995–2020	AMG, CCEMG	FD ↑ EF
Naqvi et al. [11]	APEC	1990–2017	DK, FMOLS	FD ↑ EF
Ozturk et al. [68]	South Asian	1971–2018	FMOLS, DOLS	FD ↓ EF
Saqib et al. [69]	Top ten countries with the biggest EF	1990–2109	CCEMG	FD ↑ EF
Yasin et al. [29]	BRICS	1995–2022	DK	FD ↓ EF
Natural Resource Rent—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Danish et al. [70]	BRICS	1992–2016	FMOLS, DOLS	NR ↓ EF
Ahmed et al. [32]	China	1970–2016	ARDL	NR ↑ EF NR ↑ CF
Ulucak et al. [25]	Top 15 Renewable Energy Consumption Economies	1996–2018	PSTR	NR ↑ EF
Ahmad et al. [71]	45 Resource-Rich Countries of Asia	1990–2018	POLS, DK	NR ↓ EF
Ullah et al. [72]	73 Developing Countries	1990–2016	PMG-ARDL	NR ↑ EF
Awosusi et al. [33]	India	1990–2016	ARDL	NR ↓ EF
Onifade [73]	OECD	2000–2019	Panel Quantile	NR ↑ EF
Dao et al. [74]	OECD	2009–2019	MQR	NR ↑ EF
Shittu et al. [75]	BRICS	1992–2018	FMOLS, DOLS, FE-OLS	NR ↑ EF
Human Development—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Kassouri and Altintas [41]	MENA	1990–2016	CCEMG	HD ↓ EF
Pata et al. [1]	Top Ten with Largest EF Economies	1992–2016	AMG	HD ↓ EF
Liu et al. [42]	G-7	1992–2018	CUP-FM, CUP-BC	HD ↓ EF
Qiu and Wan [43]	BRICS	1995–2019	CS-ARDL	HD ↓ EF
Balsalobre-Lorente et al. [76]	G-7	1991–2018	CUP-FM	HD ↓ EF
Nguea and Fotio [77]	31 African Countries	1996–2018	Panel Quantile	HD ↓ EF
Technological Innovation—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Sahoo and Sethi [78]	Newly Industrialized Countries	1990–2017	MG, PMG, AMG	TEC ↓ EF
Chunling et al. [79]	Pakistan	1992–2018	ARDL	TEC ↑ EF
Jahanger et al. [15]	Developing Countries	1990–2016	PMG-ARDL	TEC ↓ EF
Usman and Radulescu, [80]	highest nuclear energy-producing countries	1990–2019	AMG, CCEMG	TEC ↑ CF
Bashir et al. [81]	Newly Industrialized Countries	1990–2018	CS-ARDL	TEC ↓ EF
Dai et al. [82]	ASEAN	1995–2018	CUP-FM, CUP-BC	TEC ↓ EF
Raza et al. [37]	G-20	1990–2021	CS-ARDL	TEC ↓ EF
Chopra et al. [83]	5 high-emitting countries	1990–2022	CS-ARDL	TEC ↓ CF
Quing et al. [63]	South Asian countries	1990–2020	CCEMG	TEC ↓ EF
Nathaniel et al. [84]	Emerging Countries	2000–2020	AMG	TEC ↓ EF
Tiwari et al. [64]	USA	1990–2021	ARDL	TEC ↓ EF

Table 2. Summary of the literature (control variables).

Globalization—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Usman et al. [85]	USA	1985Q1–2014Q4	ARDL	GL ↑ EF
Omoke et al. [27]	Nigeria	1971–2014	NARDL	Short run GL ↓ CF Long run GL ↑ CF Short and long-run GL ↓ EF
Saud et al. [86]	Belt and Road	1990–2014	PMG	GL ↓ EF
Wang [47]	Brazil, Russia, India, and China	1997–2016	ARDL	GL ↓ EF GL ↓ CF
Kirikkaleli et al. [87]	Turkiye	1985–2017	FMOLS, DOLS	GL ↑ EF
Ansari et al. [88]	Top Renewable Energy-Consuming Countries	1991–2016	PMG, FMOLS DOLS	GL ↓ EF
Pata [89]	BRIC	1971–2016	FARDL	GL ↑ EF
Ehigiamusoe et al. [90]	31 African Nations	1995–2019	FMOLS	GL ↑ CO2 GL ≠ EF
Hassan et al. [48]	OECD	1990–2019	AMG, CCEMG	GL ↑ EF
Quing et al. [63]	South Asian	1990–2020	AMG, CCEMG	GL ↑ EF
Trade Openness—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Altıntas and Kassouri [41]	20 EU Countries	1985–2016	ARDL	Short run TO ↑ CF Long run TO ↓ CF
Lu [54]	13 Asian Countries	1973–2014	PMG	TO ↓ EF
Kongbuamai [91]	Thailand	1974–2016	ARDL	Short and Long run TO ↑ EF
Destek and Sinha [49]	24 Organization for Economic Co-operation and Development countries	1980–2014	MG, FMOLS, DOLS	TO ↓ EF
Aydin and Turan [92]	BRICS	1996–2016	AMG, CCEMG	TO ↓ EF
Dada et al. [52]	Nigeria	1970–2017	ARDL	TO ↑ EF
Wang et al. [51]	G-7	1990–2020	CS-ARDL	TO ↑ EF
Liu et al. [93]	Pakistan	1980–2017	ARDL	TO ↓ EF
Opuala et al. [53]	West Africa	1980–2017	PMG	TO ↑ EF
Esmaeili et al. [55]	19 Energy-Intensive Countries	1997–2018	ARDL, CS-ARDL	Short run TO ↑ EF Long run TO ↓ EF
Javed et al. [94]	Italy	1994–2019	DARDL	TO ↑ EF
Abdullahi et al. [95]	Ten ECOWAS Countries	1980–2022	PMG	TO ↑ EF
Urbanization—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Ahmed et al. [96]	G-7	1971–2014	CUP-FM, CUP-BC	UB ↑ EF
Ahmed et al. [32]	China	1970–2016	ARDL	UB ↑ EF
Nathaniel et al. [59]	CIVETS	1990–2014	AMG	UB ↑ EF
Nathaniel and Khan [97]	ASEAN	1990–2016	AMG	UB ≠ EF
Nathaniel [98]	Indonesia	1971–2014	ARDL	UB ↑ EF
Salman et al. [56]	ASEAN-4	1980–2017	ARDL	UB ≠ EF
Ponce et al. [99]	100 Countries	1980–2019	ARDL	UB ↓ CF
Shah et al. [57]	Top 15 Natural Gas Supplier Economies	2000–2019	CS-ARDL, AMG	UB ↑ EF
Hussain et al. [100]	E-7	1992–2020	FMOLS	UB ↑ EF
Aziz et al. [101]	Saudi Arabia	1991–2021	ARDL	UB ↑ EF
Mehmood [102]	G-11	1990–2020	CS-ARDL	UB ↑ EF
Renewable Energy—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Usman and Radulescu, [80]	Highest Nuclear Energy-Producing Countries	1990–2019	AMG, CCEMG	RNW ↓ CF
Saqib, [66]	63 Emerging and Developed Countries	1990–2020	AMG, CCEMG	RNW ↓ CF
Radmehr et al. [103]	EU	1995–2018	Spatial Panel	RNW ↓ EF
Rahman et al. [104]	India	1980–2021	ARDL	RNW ↓ CF
Joof et al. [105]	USA	1980–2018	ARDL	RNW ↓ EF
Sohag et al. [106]	OECD	1990–2018	CS-ARDL	RNW ↓ EF
Sethi et al. [107]	BRICS+	2000–2020	NARDL	RNW ↓ EF

Table 2. Summary of the literature (control variables).

Globalization—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Usman et al. [85]	USA	1985Q1–2014Q4	ARDL	GL ↑ EF
Omoke et al. [27]	Nigeria	1971–2014	NARDL	Short run GL ↓ CF Long run GL ↑ CF Short and long-run GL ↓ EF
Saud et al. [86]	Belt and Road	1990–2014	PMG	GL ↓ EF
Wang [47]	Brazil, Russia, India, and China	1997–2016	ARDL	GL ↓ EF GL ↓ CF
Kirikkaleli et al. [87]	Turkiye	1985–2017	FMOLS, DOLS	GL ↑ EF
Ansari et al. [88]	Top Renewable Energy-Consuming Countries	1991–2016	PMG, FMOLS DOLS	GL ↓ EF
Pata [89]	BRIC	1971–2016	FARDL	GL ↑ EF
Ehigiamusoe et al. [90]	31 African Nations	1995–2019	FMOLS	GL ↑ CO2 GL ≠ EF
Hassan et al. [48]	OECD	1990–2019	AMG, CCEMG	GL ↑ EF
Quing et al. [63]	South Asian	1990–2020	AMG, CCEMG	GL ↑ EF
Trade Openness—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Altıntas and Kassouri [41]	20 EU Countries	1985–2016	ARDL	Short run TO ↑ CF Long run TO ↓ CF
Lu [54]	13 Asian Countries	1973–2014	PMG	TO ↓ EF
Kongbuamai [91]	Thailand	1974–2016	ARDL	Short and Long run TO ↑ EF
Destek and Sinha [49]	24 Organization for Economic Co-operation and Development countries	1980–2014	MG, FMOLS, DOLS	TO ↓ EF
Aydin and Turan [92]	BRICS	1996–2016	AMG, CCEMG	TO ↓ EF
Dada et al. [52]	Nigeria	1970–2017	ARDL	TO ↑ EF
Wang et al. [51]	G-7	1990–2020	CS-ARDL	TO ↑ EF
Liu et al. [93]	Pakistan	1980–2017	ARDL	TO ↓ EF
Opuala et al. [53]	West Africa	1980–2017	PMG	TO ↑ EF
Esmaeili et al. [55]	19 Energy-Intensive Countries	1997–2018	ARDL, CS-ARDL	Short run TO ↑ EF Long run TO ↓ EF
Javed et al. [94]	Italy	1994–2019	DARDL	TO ↑ EF
Abdullahi et al. [95]	Ten ECOWAS Countries	1980–2022	PMG	TO ↑ EF
Urbanization—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Ahmed et al. [96]	G-7	1971–2014	CUP-FM, CUP-BC	UB ↑ EF
Ahmed et al. [32]	China	1970–2016	ARDL	UB ↑ EF
Nathaniel et al. [59]	CIVETS	1990–2014	AMG	UB ↑ EF
Nathaniel and Khan [97]	ASEAN	1990–2016	AMG	UB ≠ EF
Nathaniel [98]	Indonesia	1971–2014	ARDL	UB ↑ EF
Salman et al. [56]	ASEAN-4	1980–2017	ARDL	UB ≠ EF
Ponce et al. [99]	100 Countries	1980–2019	ARDL	UB ↓ CF
Shah et al. [57]	Top 15 Natural Gas Supplier Economies	2000–2019	CS-ARDL, AMG	UB ↑ EF
Hussain et al. [100]	E-7	1992–2020	FMOLS	UB ↑ EF
Aziz et al. [101]	Saudi Arabia	1991–2021	ARDL	UB ↑ EF
Mehmood [102]	G-11	1990–2020	CS-ARDL	UB ↑ EF
Renewable Energy—Ecological Footprint and Carbon Footprint
Authors	Countries/Groups and Years		Estimators	Empirical Outcomes
Usman and Radulescu, [80]	Highest Nuclear Energy-Producing Countries	1990–2019	AMG, CCEMG	RNW ↓ CF
Saqib, [66]	63 Emerging and Developed Countries	1990–2020	AMG, CCEMG	RNW ↓ CF
Radmehr et al. [103]	EU	1995–2018	Spatial Panel	RNW ↓ EF
Rahman et al. [104]	India	1980–2021	ARDL	RNW ↓ CF
Joof et al. [105]	USA	1980–2018	ARDL	RNW ↓ EF
Sohag et al. [106]	OECD	1990–2018	CS-ARDL	RNW ↓ EF
Sethi et al. [107]	BRICS+	2000–2020	NARDL	RNW ↓ EF

Research Gap

A review of the existing literature reveals some empirical gaps. (1) Most existing studies generally focus on a single variable to measure environmental degradation, and carbon emissions (CO₂) are widely used. However, while research on the ecological footprint and carbon emissions is increasing, findings on the carbon footprint are limited. In this context, this study aims to strengthen previous findings by combining ecological and carbon footprints to comprehensively analyze environmental degradation. (2) However, the impacts of financial development, natural resource rent, technological innovation, and human development on environmental sustainability show variable results across different country groups. However, research on the G-10 countries, which consist of industrialized economies, is quite limited. Examining the connection between environmental degradation and economic indicators for these countries, which play a key role in the global economy, will provide a new regional perspective to the literature. (3) Related variables have been used before, but few models examine economic, social, and environmental indicators together and focus on the asymmetric linkage between financial development and environmental degradation. To address these shortcomings, this study focuses on the effect of financial development, natural resource rent, technological innovation, and human development on ecological and carbon footprints in the context of G-10 countries. In this framework, the following hypotheses are formulated based on the theoretical and empirical discussions in the literature.

H1. 

An inverted U-shaped relationship between financial development and environmental pollution exists.

H2. 

Natural resource rent increases environmental pollution.

H3. 

Technological innovation reduces environmental pollution.

H4. 

Renewable energy reduces environmental pollution.

3. Variable Selection, Model Construction, and Methodology

3.1. Variable Selection

This research analyzes the impacts of financial development, natural resource rent, technological innovation, and human development on the carbon and ecological footprint. The variables used in this study and their theoretical foundations are presented below. In this context, using panel data, our study examines the linkage between these explanatory variables and the ecological and carbon footprints of G-10 countries.

3.1.1. Dependent Variables

This study measures environmental pressure using ecological and carbon footprints. Ecological footprints cover various human activities such as land use, water consumption, energy use, waste generation, and biocapacity [105]. In addition, the ecological footprint provides a broader perspective than the carbon footprint, considering the human demand for natural resources and ecosystems. Combining these two indicators aims to provide more holistic, comprehensive, and comparable results on environmental sustainability.

3.1.2. Independent Variables

The previous literature shows that financial development has a dual effect on environmental pollution, with some studies indicating an increase in pollution [3,10,11,30]. In contrast, others report a decrease [13,15,27] in financial development, which affects economic growth. As a result, it can trigger environmental degradation through rapid production and industrialization or promote clean energy by converting savings into investment and positively affecting environmental quality. Financial development can have scale and compositional effects on pollution in this framework. Previous studies [17,18,31] have confirmed an inverted U-shaped link between FD and pollution. Therefore, this study reanalyzes this relationship by focusing on G-10 countries. Conversely, the utilization of natural resources drives economic growth. As economies develop, these resources’ use increases, often leading to negative environmental impacts. However, sustainable management and technological innovation can reduce resource use. In addition, human development plays a role in resource use and thus affects pollution levels. Based on these discussions, we include financial development, natural resource rent, technological innovation, and human development as independent variables in the model.

3.1.3. Control Variables

G-10 countries play important roles in the global economy, leading international trade and pioneering globalization. While globalization boosts economic activity, energy demand, and financial development through foreign direct investment and trade, there is mixed evidence of its environmental impacts [86]. Some studies [48,85,89] argue that globalization promotes environmental degradation, while others [47,86,88] argue that pollution can be reduced by adopting environmentally friendly technologies. Urbanization is also an important factor in managing the environmental impacts of G-10 countries. While urbanization increases economic activity and demand for renewable energy, it can also trigger fossil fuel dependence and pollution due to industrialization and transportation [59,96]. Despite these drawbacks, urbanization also increases the spending power of urban residents, increases the demand for renewable energy, and helps environmental remediation efforts [59]. On the other hand, increased environmental awareness encourages policymakers and the industrial sector to prioritize renewable energy sources, reducing environmental degradation [49]. Empirical findings in the literature [37,63,64] agree that renewable energy improves environmental quality. Based on these discussions, globalization, trade openness, urbanization, and renewable energy consumption, which are considered to affect the sustainability of the environment, are added to the model as control variables.

3.2. Data and Model Construction

Ten different variables were selected to test the EF and CF functions and to estimate the models. These variables were collected from five databases (GFN, WB, OWID, KOF, and IMF) for the G-10 countries. The Global Footprint Network (GFN) [108] has sourced ecological and carbon footprint data. Financial development index data come from the International Monetary Fund (IMF) [109]. Globalization Index data have been gathered from the KOF [110] Swiss Economic Institute. Human development index data are available from Our World in Data (OWID) [111]. Trade Openness, Urbanization, and Natural Resources Rent data are also derived from World Bank (WB) [112] databases. For countries and the number of variables, the study period is defined as the first and last date range for which data on the indicators used to construct the panel dataset are available. The Ecological Footprint and Carbon Footprint data in the GFN database were last calculated in 2022, and the human development data in the OWID database are available for the first time since 1990. This study covers the period from 1990 to 2022, using the most extensive dataset available for the variables.

Table 3. Variable description and sources of data.

	Variable	Acronym	Definition	Sources
Dependent	Ecological Footprint	EF	Ecological Footprint (gha per person)	GFN
Dependent	Carbon Footprint	CF	Carbon footprint (gha per person)	GFN
Independent	Financial Development	FD	Financial Development Index	IMF
Independent	Financial Development	FD2	Financial Development Index	IMF
Independent	Natural Resources Rent	NR	Natural resources (% of GDP)	WB
Independent	Human Development	HD	Human Development Index	OWID
Independent	Technological Innovation	TEC	Patent applications (residents + nonresidents)	WB
Control	Renewable Energy Consumption	RNW	% of total final energy consumption	WB
Control	Trade Openness	TO	The share of total imports and exports (% of GDP)	WB
Control	Urbanization	UB	Urbanization (% of total population)	WB
Control	Globalization	GL	KOF Globalization Index	KOF

provides detailed information on measurement indicators and data sources. In 2022, the world’s average ecological footprint was announced as 2.58 gha per capita; the average carbon footprint was 1.56. During the same period, in G-10 countries, the ecological footprint’s average per capita stood at 5.149 gha, and the average carbon footprint was 2.986 gha [108]. These values suggest that environmental degradation in G-10 countries exceeds the world average. The global average of the financial development index was also reported as 0.32; in G-10 countries, it was significantly higher at 0.81 [110]. This indicates that the financial development level of G-10 countries surpasses the world average. While the world average for natural resource rent is 3, this ratio is 0.789 for the G-10 countries [112]. Although this value calculated for the G-10 is below the world average, it can be said that the ratio is generally high for this group of countries when all countries are considered. The world average for the human development index is 0.739, and for the G-10, it is 0.932 [111]. This shows that the level of human development in the G-10 countries is higher than the world average. The global average of patent applications considered for technological innovation is about one point seven million, while it is ninety-three thousand in the G-10 countries [112]. This shows that the G-10 countries have a 5.47% world share.

This research examines the link between FD, EF, and CF by following an inverted U-shaped curve and explores the long-run linkages between NR, HD, and TEC and the EF and CF variables. The control variables RNW, TO, UB, and GL, which are thought to affect carbon and ecological footprints, are also estimated in the model. These variables are analyzed using logarithmic forms to address potential multicollinearity and endogeneity issues in the models. The paper constructs two empirical models for the EF and CF functions. In this study, two fundamental econometric functions are examined. These are clarified by Equations (1) and (2).

$${\mathrm{EF}}_{\mathrm{it}}=f_1\left({\mathrm{FD}}_{\mathrm{it}},{\mathrm{FD}}_{\mathrm{it}}^2,{\mathrm{NR}}_{\mathrm{it}},{\mathrm{HD}}_{\mathrm{it}},{ \mathrm{TEC}}_{\mathrm{it}},{ \mathrm{RNW}}_{\mathrm{it}},{ \mathrm{TO}}_{\mathrm{it}},{ \mathrm{UB}}_{\mathrm{it}},{ \mathrm{GL}}_{\mathrm{it}}\right)$$

(1)

$${\mathrm{CF}}_{\mathrm{it}}=f_2\left({\mathrm{FD}}_{\mathrm{it}},{\mathrm{FD}}_{\mathrm{it}}^2,{\mathrm{NR}}_{\mathrm{it}},{\mathrm{HD}}_{\mathrm{it}},{ \mathrm{TEC}}_{\mathrm{it}},{ \mathrm{RNW}}_{\mathrm{it}},{ \mathrm{TO}}_{\mathrm{it}},{ \mathrm{UB}}_{\mathrm{it}},{ \mathrm{GL}}_{\mathrm{it}}\right)$$

(2)

Equations (3) and (4) represent the models estimated by using empirical functions. Notably, these equations apply to the G-10 countries. Here, ‘i’ denotes each country, while ‘t’ denotes time. The term ‘μ_it’ signifies the stochastic error; ‘β₀’ is the constant term; and ‘β₁–β₇’ denotes the estimated coefficients.

$$\begin{gathered} \mathrm{Ln}({\mathrm{EF}}_{\mathrm{it}})={\beta }_0+{\beta }_1\mathrm{Ln}\left({\mathrm{FD}}_{\mathrm{it}}\right)+{\beta }_2\mathrm{Ln}\left({\mathrm{FD}}_{\mathrm{it}}^2\right)+{\beta }_3\mathrm{Ln}\left({\mathrm{NR}}_{\mathrm{it}}\right)+ \\ {\beta }_4\mathrm{Ln}\left({\mathrm{HD}}_{\mathrm{it}}\right)+{\beta }_5\mathrm{Ln}\left({\mathrm{TEC}}_{\mathrm{it}}\right)+{\beta }_6\mathrm{Ln}\left({\mathrm{RNW}}_{\mathrm{it}}\right)+{\beta }_7\mathrm{Ln}\left({\mathrm{TO}}_{\mathrm{it}}\right)+{\beta }_8\mathrm{Ln}\left({\mathrm{UB}}_{\mathrm{it}}\right)+{\beta }_9\mathrm{Ln}\left({\mathrm{GL}}_{\mathrm{it}}\right)+{\mu }_{\mathrm{it}} \end{gathered}$$

(3)

$$\begin{gathered} \mathrm{Ln}({\mathrm{CF}}_{\mathrm{it}})={\beta }_0+{\beta }_1\mathrm{Ln}\left({\mathrm{FD}}_{\mathrm{it}}\right)+{\beta }_2\mathrm{Ln}\left({\mathrm{FD}}_{\mathrm{it}}^2\right)+{\beta }_3\mathrm{Ln}\left({\mathrm{NR}}_{\mathrm{it}}\right) \\ +{\beta }_4\mathrm{Ln}\left({\mathrm{HD}}_{\mathrm{it}}\right)+{\beta }_5\mathrm{Ln}\left({\mathrm{TEC}}_{\mathrm{it}}\right)+{\beta }_6\mathrm{Ln}\left({\mathrm{RNW}}_{\mathrm{it}}\right)+{\beta }_7\mathrm{Ln}\left({\mathrm{TO}}_{\mathrm{it}}\right)+{\beta }_8\mathrm{Ln}\left({\mathrm{UB}}_{\mathrm{it}}\right)+{\beta }_9\mathrm{Ln}\left({\mathrm{GL}}_{\mathrm{it}}\right)+{\mu }_{\mathrm{it}} \end{gathered}$$

(4)

These equations assume that FD and its square initially have a positive relationship with EF and CF (β₁ > 0), subsequently turning negative (β₂ < 0). Consistent with the existing literature, we assume that NR and TEC negatively affect environmental quality (β₃ > 0, β₅ > 0). Conversely, HD is assumed to reduce EF and CF in G-10 countries (β₄ < 0).

3.3. Methodology

It is essential to conduct initial tests in the panel data analysis to investigate the long-term effects of various factors on ecological and carbon footprints. These factors include financial development, human development, natural resources, and technological innovation. Neglecting these tests could result in inaccurate and inconsistent estimation results. This study utilized kurtosis, skewness, and the Jarque–Bera test statistics to assess normal distribution. Furthermore, the Spearman correlation test was employed to calculate correlation coefficients between the series. This test is a statistical technique used when the assumption of normality is unmet. The correlation coefficients are determined using Equation (5).

$${\mathrm{r}}_{\mathrm{s}}=1-{\displaystyle \frac{6{{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{d}}_{\mathrm{i}}^2}{\mathrm{n}\left({\mathrm{n}}^2-1\right)}}$$

(5)

In the following equation, X and Y can be any two variables, while i takes on values from 1 to n. The pairs (xi, yi) represent the ith measurement of these variables. This study uses tolerance and variance inflation factor (VIF) coefficients to check for multicollinearity. The VIF measures how the multicollinearity between the explanatory variables inflates the estimated regression coefficient’s variance. Both VIF and tolerance coefficients are calculated using Equation (6).

$$\mathrm{VIF}={\displaystyle \frac{1}{1-{\mathrm{R}}^2}}={\displaystyle \frac{1}{\mathrm{Tolerance}}}$$

(6)

Tests such as the exogenous Wald and Sargan–Hansen tests are utilized to detect endogeneity, which arises when the impact of the explanatory variable on the dependent variable is uncertain. Cross-sectional dependence (CSD) is crucial for obtaining reliable results. Additionally, the CSD results are essential for conducting unit root tests to ensure the reliability of stationarity. This study examined CSD for variables and models using the Bias-Corrected Scaled LM test developed by Pesaran et al. [113]. The CSD LM_adj test can be found in Equation (7).

$${\mathrm{LM}}_{adj}=\sqrt{{\displaystyle \frac{2\mathrm{T}}{\mathrm{N}\left(\mathrm{N}-1\right)}}}\left({{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{N}-1}{{\displaystyle \sum }}_{\mathrm{j}=\mathrm{i}+1}^{\mathrm{N}}\hat{\Psi }\mathrm{ij}\right) \left[{\displaystyle \frac{\left(\mathrm{T}-\mathrm{K}\right){\hat{\Psi }}_{\mathrm{ij}}^2-\left(\mathrm{T}-\mathrm{K}\right){\hat{\Psi }}_{\mathrm{ij}}^2}{\mathrm{Var}\left(\mathrm{T}-\mathrm{K}\right){\hat{\Psi }}_{\mathrm{ij}}^2}}\right]$$

(7)

Following the CSD test, it is crucial to examine the heterogeneity of the slopes of both variables and models. The delta test [114] analyzes panel data. Equation (8) shows the delta test.

$${\tilde{\Delta }}_{\mathrm{adj}}=\sqrt{\mathrm{N}}{\displaystyle \frac{{\mathrm{N}}^{-1 }\widehat{\mathrm{S} }-\mathrm{E}( \check{{\mathrm{Z}}_{\mathrm{it}} )} }{\sqrt{\mathrm{Var}} ( \check{{\mathrm{Z}}_{\mathrm{it}} )}}}~\mathrm{N}\left(0,1\right)$$

(8)

Underlining the thoroughness of the analysis, this section employs a range of unit root tests, including Levin et al.’s [115] LLC test, Im et al.’s [116] IPS test, and Hadri’s [117] first-generation unit root test, to analyze the slope heterogeneity and CSD and verify the series’ stationarity. Panel unit root tests are used for heterogeneous and homogeneous series without CDS, as illustrated in Equation (9).

$$\Delta y_{i,t}={\alpha }_i+{\beta }_{i,t}+{\theta }_t+\rho y_{i,t-1}+{{\displaystyle \sum }}_{j=1}^k{\varnothing }_k\Delta y_{i,t-j}+{\mu }_{i,t}$$

(9)

The IPS test extends the LLC test by considering variability in the coefficient of y_i,t−1, and the panel unit root test calculates the average of the individual unit root statistics using Equation (10).

$$\Delta {\mathrm{X}}_{\mathrm{it}}={\beta }_{\mathrm{i}}+{\pi }_{\mathrm{i}}{\mathrm{x}}_{\mathrm{i},\mathrm{t}-1}+{\lambda }_{\mathrm{i}}{\overline{\mathrm{x}}}_{\mathrm{t}-1}+{\delta }_{\mathrm{i}}\Delta {\overline{\mathrm{x}}}_{\mathrm{t}}+{\epsilon }_{\mathrm{it}}$$

(10)

When a lagged value (t − 1) is added, Equation (10) becomes Equation (11).

$$\Delta {\mathrm{X}}_{\mathrm{it}}={\beta }_{\mathrm{i}}+{\pi }_{\mathrm{i}}{\mathrm{x}}_{\mathrm{i},\mathrm{t}-1}+{\lambda }_{\mathrm{i}}{\overline{\mathrm{x}}}_{\mathrm{t}-1}+{{\displaystyle \sum }}_{\mathrm{j}=0}^{\mathrm{p}}{\delta }_{\mathrm{i}}\Delta {\overline{\mathrm{x}}}_{\mathrm{t}-\mathrm{j}}+{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{p}}{\varphi }_{\mathrm{ij}}{\Delta \mathrm{x}}_{\mathrm{i},\mathrm{t}-\mathrm{j}}+{\epsilon }_{\mathrm{it}}$$

(11)

When conducting the Hadri [117] test, the null hypothesis is that the series under consideration is stationary, unlike the LLC and IPS tests. The equations for the fixed effect and trend tests are provided below.

$$Y_{it}=r_{it}+{\epsilon }_{it}$$

(12)

$$Y_{it}=r_{it}+{\beta }_{i}t+{\epsilon }_{it}$$

(13)

In Equations (12) and (13), $r_{it}=r_{it-1}+u_{it}$. If the null hypothesis holds, ${\sigma }_u^2=0$. This means that r_it is constant; therefore, the Y_it series is stationary. First-generation unit root tests are employed to test the stationarity of the series. LLC and IPS tests involve null hypothesis tests for ρ = 0 and alternative tests for ρ < 0. Once it is determined that variables are integrated, testing for a long-term equilibrium linkage is the next step.

This study employs Westerlund’s [118] DH and Kao’s [119] panel cointegration tests to analyze these long-term relationships between variables. Westerlund’s [118] technique assumes that the series are stationary in their first differences and, in particular, that the dependent variable has a unit root. This method can provide reliable results under panel homogeneity. Moreover, this method does not restrict the number of variables when investigating cointegration relationships; considering the number of variables in the current study, it is suitable for examining long-run relationships. Kao’s [119] cointegration addresses and accounts for panel standard errors such as endogeneity, heteroskedasticity, and autocorrelation. In the panel data of this study, the dependent variable and other series have unit roots and are made stationary by taking their first differences. In addition, the panel data model includes homogeneity and possible standard errors. The panel data characteristics meet the necessary conditions for applying these two tests and are appropriate. Following the panel data characteristics, the findings of both methods are mutually supportive. The Kao [119] test is detailed in Equation (14).

$$\mathrm{ADF}={\mathrm{t}}_{\overline{\mathrm{P}}}+\sqrt{{\displaystyle \frac{6{\mathrm{N}\hat{\sigma }}_{\mathrm{r}}}{2{\hat{\sigma }}_{0\mathrm{r}}}}}/\sqrt{{\displaystyle \frac{{\sigma }_{0\mathrm{r}}^{\widehat{2}}}{2{\sigma }_{\mathrm{r}}^{\widehat{2}}}}}+3{\sigma }_{\mathrm{r}}^{\widehat{2}}{\mathrm{r}}_{\mathrm{r}}^2/10\left(6{\hat{\sigma }}_{0\mathrm{r}}\right)$$

(14)

The Westerlund [118] method employs two separate test statistics: the Durbin-H panel (DH_p) and the Durbin-H group (DH_g). The DH_g provides reliable results when the autoregressive parameter is heterogeneous across panels, while the DH_p is robust when the autoregressive parameter is homogeneous across panels. The DH_g and DH_p are calculated using Equations (15) and (16).

$${\mathrm{DH}}_{\mathrm{g}}={{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}{\hat{\mathrm{S}}}_{\mathrm{i}}{\left({\tilde{\varphi }}_{\mathrm{i}}-{\tilde{\varphi }}_{\mathrm{i}}\right)}^2{{\displaystyle \sum }}_{\mathrm{t}=2}^{\mathrm{T}}{\hat{\mathrm{e}}}_{\mathrm{it}}^2-1$$

(15)

$${\mathrm{DH}}_{\mathrm{P}}={\hat{\mathrm{S}}}_{\mathrm{n}}{\left(\tilde{\varphi }-\tilde{\varphi }\right)}^2{{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}{{\displaystyle \sum }}_{\mathrm{t}=2}^{\mathrm{T}}{\hat{\mathrm{e}}}_{\mathrm{it}-1}^2$$

(16)

In the equations above, S_i donates the variance, Φ stands for the cointegration parameter, and e represents the error term. Cointegration tests are a statistical tool used to determine whether two or more variables are cointegrated. These tests evaluate the null hypothesis H₀: ρ = 1 and the alternative hypothesis H₁: ρ ≠ 1 against the absence of cointegration.

This study applies panel data analysis, which allows for the simultaneous analysis of time and cross-sectional data for the G-10 countries to examine the factors affecting environmental pollution. This method investigates econometric functions constructed based on the study’s theoretical framework. Once the cointegration connection between the variables is obtained, the long-run dynamic link must be measured by estimating the coefficients. Once EF and CF and the explanatory variables are found to be cointegrated, the elasticity coefficients for the impact of FD, NR, HD, and TEC on pollution are estimated, as well as the long-run relationships between EF and CF and the control variables RNW, TO, UB, and GL. Stock and Watson’s [120] Dynamic Ordinary Least Squares (DOLS) and Pedroni’s [121] Fully Modified Ordinary Least Squares (FMOLS) are common and preferred approaches for estimating elasticity coefficients. FMOLS and DOLS techniques are applied to the main results of this study. These techniques utilize standard fixed effects estimators that adjust for biases caused by heteroskedasticity, autocorrelation, and endogeneity in panel data without CSD. In this paper, we estimate the coefficients using FMOLS and DOLS methods, which provide robust and reliable results under slope homogeneity to correct for possible panel standard errors. The FMOLS model is represented in Equation (17).

$${\hat{\beta }}_{\mathrm{FMOLS}}^{*}={\mathrm{N}}^{-1}{{\displaystyle \sum }}_{\mathrm{n}=1}^{\mathrm{N}}{\hat{\beta }}_{\mathrm{FMOLS},\mathrm{n}}^{*}$$

(17)

In the equation, ${\widehat{\beta }}_{FMOLS,n}^{*}$ represents FMOLS estimates for cross-sections. The t-statistics are calculated via Equation (18).

$${\mathrm{t}}_{{\hat{\beta }}_{\mathrm{FMOLS}}}={\mathrm{N}}^{-1/2}{{\displaystyle \sum }}_{\mathrm{n}=1}^{\mathrm{N}}{\mathrm{t}}_{{\beta }_{\mathrm{FMOLS}},\mathrm{n}}$$

(18)

where $t_{{\widehat{\beta }}_{FMOLS}}$ denotes the t-statistics of the FMOLS.

$${\mathrm{y}}_{\mathrm{i},\mathrm{t}}={\alpha }_{\mathrm{i}}+{\beta }_{\mathrm{i}}{\mathrm{x}}_{\mathrm{i},\mathrm{t}}+{{\displaystyle \sum }}_{\mathrm{j}=-\mathrm{q}}^{\mathrm{q}}{\mathrm{c}}_{\mathrm{ij}}\Delta {\mathrm{x}}_{\mathrm{it}+\mathrm{j}}+{\mathrm{e}}_{\mathrm{i},\mathrm{t}}$$

(19)

Equation (19) displays the DOLS estimator.

As an additional check to complement the results and ensure consistency and validity, the robustness of the findings was checked by applying two specific techniques, M and S, to correct standard errors. These methods are robust enough to overcome panel standard errors and provide consistent and efficient results in models with multifactor variables and different degrees of integration. To ensure the robustness and reliability of the forecasts, Breusch–Pagan–Godfrey tests were conducted for heteroskedasticity, Breusch–Godfrey LM tests were conducted for serial correlation, Jarque–Bera tests were conducted for normal distribution, and Ramsey’s Reset test was conducted for model specification.

This study analyzed these relationships using the Emirmahmutoglu and Kose (EK) [122] panel bootstrap causality test. This method is a next-generation panel causality approach that extends causality testing to panel data, minimizes unit root and cointegration bias, and delivers robust results in the absence of CSD in homogeneous panel datasets. The panel data of the study are appropriate for E-K causality results. This test calculates Fisher statistics using the LA-VAR approach and can be used in heterogeneous panels. The EK test is versatile; it works at different stationarity levels and does not worry about whether variables are cointegrated. In the vector autoregressive (VAR) model for heterogeneous panels (ki + dmax) H₀ = A_1,2,i,j = 0, the estimated hypothesis is calculated as shown in Equation (20).

$${\mathrm{z}}_{\mathrm{i},\mathrm{t}}={\mu }_{\mathrm{i}}+{\mathrm{A}}_{\mathrm{i}1}{\mathrm{z}}_{\mathrm{i},\mathrm{t}-1}+\dots \dots +{\mathrm{A}}_{{\mathrm{ik}}_{\mathrm{i}}}{\mathrm{z}}_{\mathrm{i},\mathrm{t}-{\mathrm{k}}_{ \mathrm{i} \prime }}+{{\displaystyle \sum }}_{1={\mathrm{k}}_{\mathrm{i}}+1}^{{\mathrm{k}}_{\mathrm{i}}+{\mathrm{d} \max }_{\mathrm{i}}}{\mathrm{A}}_{\mathrm{i}1}{\mathrm{z}}_{\mathrm{i},\mathrm{t}-1}+{\mathrm{u}}_{\mathrm{i},\mathrm{t}}$$

(20)

The Fisher test statistics, shown in Equation (21), are utilized in such panels.

$$ƛ=-2{{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{N}}\ln ({\mathrm{p}}_{\mathrm{i}}) \mathrm{i}=1,2,\dots ,\mathrm{N}$$

(21)

The Fisher statistic is calculated using the bootstrap method to generate an empirical distribution. Equations (22) and (23) detail this process for estimating the VAR model with k_i + dmax lags.

$${\mathrm{x}}_{\mathrm{i},\mathrm{t}}={\mu }_{\mathrm{i}}^{\mathrm{x}}+{{\displaystyle \sum }}_{\mathrm{j}=1}^{{\mathrm{k}}_{\mathrm{i}}+{\mathrm{d} \max }_{\mathrm{i}}}{\mathrm{A}}_{11,\mathrm{ij}}{\mathrm{x}}_{\mathrm{i},\mathrm{t}-\mathrm{j}}+{{\displaystyle \sum }}_{\mathrm{j}=1}^{{\mathrm{k}}_{\mathrm{i}}+{\mathrm{d} \max }_{\mathrm{i}}}{\mathrm{A}}_{12,\mathrm{ij}}{\mathrm{y}}_{\mathrm{i},\mathrm{t}-\mathrm{j}}+{\mathrm{u}}_{\mathrm{i},\mathrm{t}}^{\mathrm{x}}$$

(22)

$${\mathrm{y}}_{\mathrm{i},\mathrm{t}}={\mu }_{\mathrm{i}}^{\mathrm{y}}+{{\displaystyle \sum }}_{\mathrm{j}=1}^{{\mathrm{k}}_{\mathrm{i}}+{\mathrm{d} \max }_{\mathrm{i}}}{\mathrm{A}}_{21,\mathrm{ij}}{\mathrm{x}}_{\mathrm{i},\mathrm{t}-\mathrm{j}}+{{\displaystyle \sum }}_{\mathrm{j}=1}^{{\mathrm{k}}_{\mathrm{i}}+{\mathrm{d} \max }_{\mathrm{i}}}{\mathrm{A}}_{22,\mathrm{ij}}{\mathrm{y}}_{\mathrm{i},\mathrm{t}-\mathrm{j}}+{\mathrm{u}}_{\mathrm{i},\mathrm{t}}^{\mathrm{y}}$$

(23)

Research has indicated that macroeconomic factors respond more strongly to negative changes than positive ones. Thus, exploring potential asymmetric effects in causal relationships between these variables is vital. In addition to the E-K test, we applied the Hatemi-J [123] asymmetric causality analysis, which is an improved version of Granger and Yoon’s [124] causality analysis to reveal the asymmetric effects of both negative and positive shocks, which can confirm the inverted U-shaped relationship by considering the causality between positive and negative shocks between EF and CF and FD. Assuming two variables are integrated, Equations (24) and (25) express each one accordingly.

$${\mathrm{y}}_{\mathrm{i}1,\mathrm{t}}={\mathrm{y}}_{\mathrm{i}1,\mathrm{t}-1}+{\mathrm{e}}_{\mathrm{i}1,\mathrm{t}}={\mathrm{y}}_{\mathrm{i}1,0}+{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{t}}{\mathrm{e}}_{\mathrm{i}1,\mathrm{j}}$$

(24)

$${\mathrm{y}}_{\mathrm{i}2,\mathrm{t}}={\mathrm{y}}_{\mathrm{i}2,\mathrm{t}-1}+{\mathrm{e}}_{\mathrm{i}1,\mathrm{t}}={\mathrm{y}}_{\mathrm{i}2,0}+{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{t}}{\mathrm{e}}_{\mathrm{i}2,\mathrm{j}}$$

(25)

The representation of negative and positive shocks is detailed in Equations (26) and (27).

$${\mathrm{e}}_{\mathrm{i}1,\mathrm{t}}^{+}=\max \left({\mathrm{e}}_{\mathrm{i}1,\mathrm{t}},0\right) { \mathrm{e}}_{\mathrm{i}2,\mathrm{t}}^{+}= \max \left({\mathrm{e}}_{\mathrm{i}2,\mathrm{t}},0\right)$$

(26)

$${\mathrm{e}}_{\mathrm{i}1,\mathrm{t}}^{-}=\max \left({\mathrm{e}}_{\mathrm{i}1,\mathrm{t}},0\right) { \mathrm{e}}_{\mathrm{i}2,\mathrm{t}}^{-}= \max \left({\mathrm{e}}_{\mathrm{i}2,\mathrm{t}},0\right)$$

(27)

Based on this information, cumulative positive and negative shocks of each series are expressed as$y_{i1,t}^{+}$, $y_{i2,t}^{+}$, $y_{i1,t}^{-}$, and $y_{i2,t}^{-}$, and are defined in Equations (28)–(31).

$${\mathrm{y}}_{\mathrm{i}1,\mathrm{t}}^{+}={\mathrm{y}}_{\mathrm{i}1,0}^{+}+{\mathrm{e}}_{\mathrm{i}1,\mathrm{t}}^{+}={\mathrm{y}}_{\mathrm{i}1,0}+{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{t}}{\mathrm{e}}_{\mathrm{i}1,\mathrm{j}}^{+}$$

(28)

$${\mathrm{y}}_{\mathrm{i}2,\mathrm{t}}^{+}={\mathrm{y}}_{\mathrm{i}2,0}^{+}+{\mathrm{e}}_{\mathrm{i}2,\mathrm{t}}^{+}={\mathrm{y}}_{\mathrm{i}2,0}+{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{t}}{\mathrm{e}}_{\mathrm{i}2,\mathrm{j}}^{+}$$

(29)

$${\mathrm{y}}_{\mathrm{i}1,\mathrm{t}}^{-}={\mathrm{y}}_{\mathrm{i}1,0}^{-}+{\mathrm{e}}_{\mathrm{i}1,\mathrm{t}}^{-}={\mathrm{y}}_{\mathrm{i}1,0}+{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{t}}{\mathrm{e}}_{\mathrm{i}1,\mathrm{j}}^{-}$$

(30)

$${\mathrm{y}}_{\mathrm{i}2,\mathrm{t}}^{-}={\mathrm{y}}_{\mathrm{i}2,0}^{-}+{\mathrm{e}}_{\mathrm{i}2,\mathrm{t}}^{-}={\mathrm{y}}_{\mathrm{i}2,0}+{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{t}}{\mathrm{e}}_{\mathrm{i}2,\mathrm{j}}^{-}$$

(31)

The next step tests causality relations between these positive and negative shocks with a p-lag VAR model. Assuming that positive shocks are described as $y_t^{+}=\left(y_{1t}^{+},y_{2t}^{+}\right)$, this causality test uses the p-lag VAR model as laid out in Equation (32).

$${\mathrm{y}}_{\mathrm{t}}^{+}=\mathrm{v}+{\mathrm{A}}_1{\mathrm{y}}_{\mathrm{t}-1}^{+}+\dots \dots +{\mathrm{A}}_{\mathrm{p}}{\mathrm{y}}_{\mathrm{t}-1}^{+}+{\mathrm{u}}_{\mathrm{t}}^{+}$$

(32)

In panel data analysis, we examine the alternative hypothesis against the null hypothesis ${\beta }_{i2,r}=0,$ which states that $y_{i2,t}^{+}$ is not causing ${\mathrm{y}}_{\mathrm{i}1,\mathrm{t}}^{+}$ for cross-section unit i, it is tested using the Wald statistic ${\beta }_{i2,r}< 0$. Figure 1 illustrates all test procedures used in the present study’s panel data applied for EF and CF functions.

4. Empirical Results

4.1. Initial Statistics

Table 4. Descriptive statistics.

Box plots
Stats.	EF	CF	FD	NR	HD	TEC	RNW	TO	UB	GL
Mean	6.290	3.859	0.749	0.519	73.041	84022	12.686	0.893	80.517	82.291
Median	5.851	3.600	0.759	0.118	62.424	16533	9.100	0.897	79.234	83.143
Max.	10.927	7.853	1.000	5.565	193.033	621453	57.900	0.967	98.153	90.929
Min.	3.461	1.910	0.354	0.008	15.723	617	0.600	0.780	66.706	55.636
Std. Dev.	1.619	1.179	0.139	0.868	39.225	151958.7	11.842	0.037	7.886	6.673
Skew.	0.804	1.171	−0.574	3.044	0.760	2.032	1.612	−0.603	0.617	−1.195
Kurt.	3.047	4.404	2.962	13.730	2.810	5.910	5.579	3.044	2.904	4.642
Jarque–Bera	39.153 ***	112.714 ***	19.958 ***	2301.673 ***	35.483 ***	377.985	257.841 ***	21.992 ***	23.149 ***	127.101 ***
Obs.	363	363	363	363	363	363	363	363	363	363
Scatters

Note: *** denotes the significance of 1%.

presents the descriptive statistics of the panel data variables used in the analysis, calculated in their raw form without logarithmic transformation. These statistics are crucial for determining whether the selected variables are exogenous. The table includes the main and Jarque–Bera (J-B) statistics for the G-10 countries. The mean values for the dependent variables EF and CF are calculated as 6.290 and 3.859 for these countries in the period under consideration. The mean values for the main explanatory variables FD, NR, HD, and TEC are 0.749, 0.519, 73.041, and 84022, respectively. TEC has the highest standard deviation, while TO has the lowest. Upon examining skewness, the EF, CF, TEC, RNW, UB, HD, and NR series are skewed to the left, whereas the FD, GL, and TO are skewed to the right. The kurtosis coefficient indicates that all series have a pointed distribution. The results of the Jarque–Bera test reject the null hypothesis, indicating that these series do not adhere to normal distribution. Box plots and scatter plots, constructed using the logarithmic form of the series, show the distribution, quarterly averages, and fluctuations in a panel dataset for the G-10 countries. They show that the series is randomly distributed.

Checks are conducted for endogeneity and multicollinearity between the variables used in the estimated models, as these issues can impact the consistency and accuracy of the model estimation results. To address these problems, we conducted a VIF test for multicollinearity (

Table 5. Heat plot of correlation matrix.

1/VIF	VIF	EF	1.000	0.853	−0.100	0.465	−0.087	−0.027	−0.123	0.107	0.245	−0.089
		CF	0.853	1.000	0.124	0.320	−0.001	0.159	−0.287	0.022	0.178	−0.169
0.358	2.792	FD	−0.100	0.124	1.000	0.121	0.531	0.175	0.219	−0.011	0.023	0.289
0.262	3.821	NR	0.465	0.320	0.121	1.000	0.104	0.280	0.118	−0.113	0.131	−0.032
0.250	3.998	HD	−0.087	−0.001	0.531	0.104	1.000	−0.037	0.411	0.424	0.424	0.648
0.268	3.727	TEC	−0.027	0.159	0.175	0.280	−0.037	1.000	−0.154	−0.795	−0.106	−0.592
0.664	1.507	RNW	−0.123	−0.287	0.219	0.118	0.411	−0.154	1.000	0.235	−0.017	0.278
0.231	4.325	TO	0.107	0.022	−0.011	−0.113	0.424	−0.795	0.235	1.000	0.264	0.789
0.516	1.938	UB	0.245	0.178	0.023	0.131	0.424	−0.106	−0.017	0.264	1.000	0.328
0.858	1.165	GL	−0.089	−0.169	0.289	−0.032	0.648	−0.592	0.278	0.789	0.328	1.000
Mean VIF	(2.909)		EF	CF	FD	NR	HD	TEC	RNW	TO	UB	GL

). Furthermore, a heat plot was constructed for the correlation coefficients between the series; the highest correlation is 0.789 between GL and TO, and the correlation levels between the other series are optimal for model estimation. We also performed Block exogenous Wald and Sargan tests for endogeneity (

Table 6. Endogeneity test results.

Sargan–Hansen Test for Exogeneity of Instruments
Instrument Specification:		Instrument Validity		Sargan–Hansen J Statistic	Prob (J-Statistic)
@DYN(EF,−2) FD(−1) FD2(−1) NR(−1) HD(−1) TEC(−1) RNW(−1) TO(−1) UB(−1) GL(−1)		Model A		1.462	0.226
@DYN(CF,−2) FD(−1) FD2(−1) NR(−1) HD(−1) TEC(−1) RNW(−1) TO(−1) UB(−1) GL(−1)		Model B		2.533	0.111
H0: The instruments used in this model are valid
Block exogenous Wald test
Hypothesis—H0: Exogenous			X2(1)			Prob.
FD	GL		2.327			0.312
	TO		0.219			0.896
	HD		0.969			0.616
	UB		2.211			0.331
	NR		0.488			0.784
	RNW		0.858			0.651
	TEC		3.300			0.192
NR	FD		0.420			0.517
	GL		0.180			0.672
	TO		0.585			0.444
	HD		0.963			0.326
	UB		0.190			0.663
	RNW		2.451			0.118
	TEC		1.266			0.261
HD	FD		0.670			0.413
	GL		2.451			0.118
	TO		0.008			0.929
	UB		0.298			0.585
	NR		0.472			0.492
	RNW		0.136			0.713
	TEC		0.077			0.781
TEC	FD		4.279			0.118
	GL		0.205			0.903
	TO		1.597			0.450
	HD		1.916			0.384
	UB		2.046			0.360
	NR		2.953			0.228
	RNW		1.874			0.392
RNW	FD		0.288			0.866
	GL		3.485			0.175
	TO		1.358			0.507
	HD		0.678			0.713
	UB		4.533			0.104
	NR		1.079			0.583
	TEC		2.149			0.341
TO	FD		0.191			0.662
	GL		0.155			0.694
	HD		0.501			0.479
	UB		1.248			0.264
	NR		2.533			0.112
	RNW		0.797			0.372
	TEC		0.618			0.432
UB	FD		0.119			0.730
	GL		2.077			0.150
	TO		0.154			0.694
	HD		0.026			0.872
	NR		0.032			0.857
	RNW		0.162			0.687
	TEC		0.681			0.409
GL	FD		0.947			0.331
	TO		1.270			0.260
	HD		1.082			0.298
	UB		0.134			0.714
	NR		1.092			0.296
	RNW		0.043			0.835
	TEC		0.253			0.615

). A VIF value of less than five indicates no multicollinearity among the variables [125,126]. In our assessment of multicollinearity, we applied the VIF test and found an average VIF of 2.909 in G-10 countries, indicating no multicollinearity issue among the variables. These findings are consistent with the patterns observed in the scatter plots.

The estimation results show no endogeneity issues in the explanatory variables when using the panel least two-stage squares model to test for endogeneity. Both the Block exogenous Wald test and the Sargan–Hansen test support the accuracy of the model’s estimation. The main and control instrumental variables selected in the research are exogenous and valid, and the models have no endogeneity (Table 6).

In international commercial and financial relations, technology is eliminating borders. As trade barriers decrease, countries are becoming more interconnected. Despite being located on different continents, the G-10 countries share similar characteristics and can be considered homogeneous. It is essential to research the relationship between different sections and variations in slope when analyzing panel data to estimate elasticity coefficients and understand the unit root process accurately. Based on the outcomes of CSD and delta tests for variables and models, we are unable to reject the null hypothesis that there is no CSD in all series, indicating no CSD in these series and models. The delta test shows that slope heterogeneity is rejected for EF, CF, FD, TEC, RNW, and GL and their corresponding models, proving that slope heterogeneity applies to TO, HD, UB, and NR (

Table 7. Slope heterogeneity and CSD results.

	Bias-Corrected LM		Delta Tests
	Stat.	p Value	$\tilde{\mathbf{\Delta }}$	p Value	${\tilde{\mathbf{\Delta }}}_{\mathit{a}\mathit{d}\mathit{j}}$	p Value
EF	0.075	0.470	1.091	0.138	1.144	0.126
CF	1.227	0.110	0.770	0.221	0.808	0.210
FD	−0.251	0.599	0.271	0.393	0.284	0.388
NR	−0.243	0.596	2.788	0.003	2.924	0.002
HD	0.947	0.172	3.931	0.000	4.123	0.000
TEC	−0.333	0.630	1.104	0.135	1.158	0.123
RNW	−0.189	0.575	−0.495	0.690	−0.519	0.698
TO	1.049	0.147	1.297	0.097	1.360	0.087
UB	0.800	0.212	18.613	0.000	19.521	0.000
GL	0.221	0.412	0.310	0.378	0.325	0.373
Model A	0.293	0.385	0.312	0.377	1.082	0.140
Model B	0.970	0.166	0.098	0.461	0.212	0.416

). First-generation unit root tests based on delta and CSD results are used to determine the stationarity levels in a series.

Based on the slope heterogeneity and CSD results, assessing the stationarity levels within the series is essential. Examining the long-term relationships between the variables and deciding which tests are appropriate is necessary. The IPS, LLC, and Hadri tests were used to check for stationarity in cases without CSD, considering both slope homogeneity and heterogeneity. After reviewing the outcomes of the unit root tests, the probability values obtained indicate that the null hypothesis suggesting a unit root within the series cannot be rejected for all series. As a result, each series is made stationary by taking their first-order differences. Consequently, all variable series are classified as I(1) rather than I(0) (

Table 8. Results of the unit root test.

Variables	Intercept						Intercept and Trend
	IPS		LLC		Hadri		IPS		LLC		Hadri
	W Stat	p Value	t Stat	p Value	Z Stat	p Value	W Stat	p Value	t Stat	p Value	Z Stat	p Value
EF	3.095	0.999	1.838	0.967	8.587	0.000	−0.044	0.482	0.081	0.532	9.237	0.000
ΔEF	−16.587	0.000	−15.990	0.000	−0.330	0.629	−14.733	0.000	−4.627	0.000	−1.118	0.868
CF	4.194	1.000	2.883	0.998	7.086	0.000	2.142	0.984	−0.522	0.301	9.831	0.000
ΔCF	−16.509	0.000	−18.772	0.000	−0.170	0.567	−14.340	0.000	−8.457	0.000	−0.508	0.694
FD	−0.412	0.340	1.329	0.908	8.648	0.000	0.408	0.658	1.648	0.950	9.519	0.000
ΔFD	−10.612	0.000	−9.164	0.000	0.978	0164	−12.605	0.000	−11.218	0.000	0.458	0.323
NR	−0.502	0.307	0.616	0.731	3.636	0.000	0.345	0.635	2.601	0.995	4.475	0.000
ΔNR	−15.566	0.000	−13.108	0.000	−0.426	0.665	−14.249	0.000	−12.031	0.000	−0.131	0.552
HD	−1.128	0.129	−0.501	0.308	11.899	0.000	0.995	0.840	1.059	0.855	8.620	0.000
ΔHD	−11.925	0.000	−12.556	0.000	−1.283	0.901	−11.521	0.000	−11.707	0.000	−0.068	0.527
TEC	−1.027	0.152	−0.507	0.305	8.839	0.000	−0.066	0.473	−0.698	0.242	7.462	0.000
ΔTEC	−8.378	0.000	−4.522	0.000	1.026	0.152	−7.187	0.000	−2.779	0.000	1.160	0.122
RNW	3.581	0.999	−0.564	0.286	10.287	0.000	−0.076	0.469	−0.262	0.396	6.700	0.000
ΔRNW	−14.902	0.000	−10.750	0.000	1.147	0.125	−14.402	0.000	−11.898	0.000	0.931	0.175
TO	2.513	0.994	−0.236	0.406	10.478	0.000	0.251	0.599	−0.108	0.457	5.503	0.000
ΔTO	−12.440	0.000	−11.225	0.000	−1.301	0.903	−10.428	0.000	−9.515	0.000	−0.125	0.549
UB	2.895	0.998	0.393	0.653	10.637	0.000	0.222	0.588	2.199	0.986	6.947	0.000
ΔUB	−8.779	0.000	−2.295	0.011	0.226	0.410	−7.052	0.000	−9.022	0.000	0.893	0.185
GL	1.049	0.853	0.089	0.535	10.412	0.000	6.998	1.000	5.305	1.000	10.244	0.000
ΔGL	−6.396	0.000	−1.884	0.029	1.140	0.127	−11.159	0.000	−4.342	0.000	1.223	0.110

). Exploring the long-run cointegration connection between variables comes after achieving stationarity.

In the preliminary panel data analysis, it has been found that neither Model A nor Model B shows slope heterogeneity or CSD. Also, all the series are stationary at I(1) levels. Following the pre-test findings, the Kao and Westerlund DH techniques were applied to examine the cointegration connection between the series. The findings of the panel cointegration tests for Model A and Model B show that both the functions EF = f₁(FD, FD², NR, HD, TEC, RNW, TO, UB, GL) and CF = f₂(FD, FD², NR, HD, TEC, RNW, TO, UB, GL) have a cointegration linkage between the series, as presented in

Table 9. Cointegration results.

Westerlund (2008) DH (Durbin–Hausman)
Model A	Value	p-Value	Model B	Value	p-Value
DHg	−2.100	0.018	DHg	−2.387	0.008
DHp	−1.679	0.047	DHp	−2.032	0.021
Kao Residual
Model A	t-stat.	prob.	Model B	t-stat.	prob.
ADF	1.846	0.032	ADF	−2.529	0.005
Residual var.	0.002		Residual var.	0.002
HAC var.	0.001		HAC var.	0.002

.

4.2. Main Results

Once the preconditions for panel estimations are determined, the elasticity coefficient estimates between variables are analyzed using FMOLS and DOLS. Moreover, M and S estimators are applied to robustify the estimation results.

Table 10. Estimation results.

Regressor	FMOLS				DOLS
	Model A 1		Model B 2		Model A 1		Model B 2
	Coef.	Prob.	Coef.	Prob.	Coef.	Prob.	Coef.	Prob.
FD	1.653	0.000	2.242	0.000	0.621	0.000	0.577	0.000
FD^2	−0.918	0.000	−1.108	0.000	−0.274	0.021	−0.663	0.000
NR	0.056	0.000	0.065	0.000	0.085	0.000	0.086	0.000
HD	−1.155	0.000	−2.004	0.000	−2.666	0.000	−7.677	0.000
TECt-1	0.017	0.010	0.042	0.000	0.085	0.000	0.042	0.013
RNW	−0.080	0.000	−0.105	0.000	−0.069	0.000	−0.045	0.002
TO	−0.045	0.002	−0.053	0.000	−0.265	0.000	−0.173	0.000
UB	0.093	0.034	0.333	0.000	0.409	0.000	0.702	0.000
GL	−0.445	0.000	−0.278	0.000	−0.375	0.011	−1.224	0.000
	Inverted U-shaped		Inverted U-shaped		Inverted U-shaped		Inverted U-shaped
Adj. R2	0.892 ***		0.861 ***		0.947 ***		0.981 ***
Jarque Bera	3.555 (0.168)		0.403 (0.817)		2.180 (0.336)		0.569 (0.752)
Ramsey’s Reset	0.921 (0.357)		1.047 (0.295)		1.384 (0.138)		1.681 (0.118)
LM	1.443 (0.124)		1.568 (0.114)		1.541 (0.118)		1.583 (0.109)
BPG	1.820 (0.178)		1.404 (0.216)		1.536 (0.184)		1.349 (0.260)
Regressor	M-estimation				S-estimation
	Model A 1		Model B 2		Model A 1		Model B 2
	coef.	prob.	coef.	prob.	coef.	prob.	coef.	prob.
FD	0.338	0.000	1.195	0.000	0.552	0.000	1.540	0.000
FD^2	−0.063	0.003	−0.304	0.000	−0.673	0.000	−0.827	0.000
NR	0.076	0.000	0.061	0.000	0.005	0.000	0.068	0.000
HD	−1.543	0.000	−1.785	0.000	−1.359	0.000	−2.607	0.000
TECt-1	0.010	0.000	0.062	0.000	0.041	0.004	0.047	0.005
RNW	−0.015	0.000	−0.050	0.000	−0.048	0.000	−0.088	0.000
TO	−0.140	0.000	−0.317	0.000	−0.075	0.000	−0.182	0.003
UB	0.910	0.000	0.943	0.000	0.445	0.000	1.595	0.000
GL	−0.707	0.000	−1.285	0.000	−0.058	0.000	−0.863	0.000
	Inverted U-shaped		Inverted U-shaped		Inverted U-shaped		Inverted U-shaped
Adj. R2	0.433 ***		0.427 ***		0.457 ***		0.208 ***
Jarque Bera	4.352 (0.113)		0.136 (0.934)		2.777 (0.249)		2.871 (0.237)
Ramsey’s Reset	1.272 (0.203)		1.434 (0.165)		1.231 (0.293)		1.448 (0.152)
LM	1.420 (0.159)		1.982 (0.143)		1.143 (0.333)		1.953 (0.154)
BPG	2.360 (0.122)		2.104 (0.125)		1.735 (0.177)		2.647 (0.104)

*** denotes the significance of 1%. ¹ ${\mathrm{EF}}_{\mathrm{it}}=f_1\left({\mathrm{FD}}_{\mathrm{it}},{\mathrm{FD}}_{\mathrm{it}}^2,{\mathrm{NR}}_{\mathrm{it}},{\mathrm{HD}}_{\mathrm{it}},{ \mathrm{TEC}}_{\mathrm{it}},{ \mathrm{RNW}}_{\mathrm{it}},{ \mathrm{TO}}_{\mathrm{it}},{ \mathrm{UB}}_{\mathrm{it}},{ \mathrm{GL}}_{\mathrm{it}}\right)$, ² ${\mathrm{CF}}_{\mathrm{it}}=f_2\left({\mathrm{FD}}_{\mathrm{it}},{\mathrm{FD}}_{\mathrm{it}}^2,{\mathrm{NR}}_{\mathrm{it}},{\mathrm{HD}}_{\mathrm{it}},{ \mathrm{TEC}}_{\mathrm{it}},{ \mathrm{RNW}}_{\mathrm{it}},{ \mathrm{TO}}_{\mathrm{it}},{ \mathrm{UB}}_{\mathrm{it}},{ \mathrm{GL}}_{\mathrm{it}}\right).$

displays long-term relationships between variables calculated by these estimators. There are no significant differences between FMOLS and DOLS regarding the direction and strength of the relationships. M and S estimators also corroborate the main results. All models show statistical significance.

The findings from FMOLS and DOLS reveal that the impact of FD and FD² on EF and CF are positive and negative, respectively. This pattern confirms the inverted U-shaped link between FD and environmental degradation (β1 > 0; β2 < 0). This finding is in line with Ashraf et al. [17] for the global sample, Khan et al. [31] for APEC, and Sun et al. [18] for South Asian nations. These results suggest that FD has scale and structural change effects on the EF and CF of G-10 countries. G-10 countries are characterized by their advanced banking systems, central banks, and capital markets, positioning them as important contributors to the global financial system. However, the findings of this study suggest that there is a divergence between financial and environmental objectives in these economies. Although financial development in these countries initially has a negative effect on the quality of the environment, after reaching a certain threshold, environmental degradation decreases. The countries in the group have undergone a period of significant economic growth. However, it can be stated that this rapid growth has a negative impact on environmental quality by increasing the use of natural resources in the initial stages of industrialization. However, after a certain threshold, it is seen that the financial sector will reduce environmental degradation by encouraging sustainable investments and increasing renewable energy projects and environmental technologies. In this context, it is essential for these countries to encourage green investments with sustainable financial policies.

On the other hand, the findings show that NR increases EF and CF (β3 > 0). This finding is in line with the findings of Shitttu et al. [75] for BRICS, and Onifade et al. [73] and Dao et al. [74] for OECD. This positive relationship between natural resource rent and environmental degradation suggests that natural resource abundance may turn into a curse in terms of ensuring environmental sustainability. Although natural resource rent has an important place in a country’s economy, it is argued that the inequality in its distribution encourages favoritism and increases the risk of corruption. That means the easy extraction of income from natural resources leads to increased corruption and patronage. This makes it difficult to allocate natural resources transparently and efficiently, negatively affecting environmental sustainability. Moreover, inequitable resource allocation and the consumption of more than nature produces threaten sustainability by creating an environmental deficit [19,34,35]. Our findings support these views. Therefore, it can be said that high natural resource rents in G-10 countries encourage the development of environmentally damaging sectors and lack sustainable natural resource management. Although in some studies [24,127], G-10 countries are often associated with good governance mechanisms and sustainable natural resource management; our findings suggest that, as Gyamfi et al. [128] noted for G-7 countries, natural resource revenues are largely directed to productivity-enhancing areas, which may further deepen environmental degradation. Therefore, it is crucial to implement country-specific policies considering the differences in natural resource rents across the G-10 countries. Consequently, G-10 countries need to adopt sustainable management practices, increase environmental taxes, and reduce energy-intensive resource use through sustainable investments.

Moreover, this study confirms a negative relationship between HD, EF, and CF (β4 < 0). This finding is in line with the findings of Liu et al. [42] and Balsalobre-Lorente et al. [76] for G-7 countries, and Qiu and Wan [43] for BRICS nations. For the G-10 countries, this suggests that human welfare and development are linked to environmental improvements. G-10 countries represent developed countries with high human development values. Sweden and Switzerland rank especially high in the human development index, while other countries show positive indicators. Therefore, it can be said that training human capital in these countries promotes sustainable development. On the other hand, our findings show that TEC increases EF and CF (β5 > 0). This finding is in line with the finding of Usman and Radulescu [80] for the highest nuclear energy-producing countries. The authors argue that a high industrial capacity and accumulation of human capital in developed countries stimulate the development of technological innovation, which is directly related to output growth, and that technological development may increase environmental degradation as a consequence of the industrialization process. Our findings support this view. Therefore, it can be stated that economic growth in G-10 countries is supported by production based on advanced technology. It is of great importance for these countries to harmonize technological innovations with environmental policies.

On the other hand, globalization, trade openness, urbanization, and renewable energy sources, which are considered to affect environmental degradation, were added to the model as control variables. RNW, CTR, and GL are found to decrease CF and EF (β6 < 0; β7 < 0; β9 < 0), whereas UB is found to increase CF and EF (β8 > 0). These findings are consistent with Wang et al. [47] for BRICS countries, Ansari et al. [88] for top renewable energy-consuming countries, Liu et al. [93] for Pakistan, Joof et al. [105] for the USA, Sohag et al. [106] for OECD, and Sethi et al. [107] for BRICS countries. Therefore, it can be said that the technical effect of globalization and trade openness is valid in these countries and environmental quality has improved through clean technology transfer and sustainable trade agreements. On the other hand, renewable energy consumption in these countries has the potential to improve environmental quality. G-10 countries are investing heavily in renewable energy sources, and countries such as Germany, Sweden, and Canada are implementing energy transition policies. Therefore, sustainable development can be supported by strengthening green energy policies in these countries.

On the other hand, urbanization, which is a dynamic socio-economic transformation process in G-10 countries, and population concentration due to rural–urban migration, is observed to increase environmental degradation. This finding is in line with the findings of Ahmed et al. [96] for G-7 countries and Mehmood et al. [102] for G-11 countries. G-10 countries have developed economies and high urbanization rates. This implies that UB in G-10 countries contributes to environmental degradation through industrialization, housing, and transport. Therefore, it is important for a sustainable environment that urban residents in G-10 countries increase their purchasing power to encourage them to choose renewable energy and invest more in green technologies and low-carbon urban projects (Figure 2).

A long-lasting relationship between variables implies that there should be causality in at least one direction [129]. Although the signs of elasticity estimates confirm the existence of an inverted U-shaped linkage for the G-10 countries, it is imperative to investigate causal links as well. The long-run results identify the factors affecting environmental pollution in the G-10 nations, but the causal direction of the relationship is equally essential in suggesting useful policies. For this purpose, the present study employs E-K panel Fisher and Hatemi-J asymmetric causality tests. The empirical results are reported in

Table 11. Results of panel Fisher causality test.

Causality			Panel Fisher Stat.	Asymptotic Prob.
FD	→	EF	76.666	0.000
NR	→	EF	66.190	0.000
HD	→	EF	68.364	0.000
TEC	→	EF	48.663	0.001
RNW	→	EF	96.092	0.000
TO	→	EF	73.093	0.000
UB	→	EF	66.983	0.000
GL	→	EF	59.021	0.000
FD	→	CF	76.474	0.000
NR	→	CF	81.927	0.000
HD	→	CF	57.663	0.000
TEC	→	CF	52.202	0.000
RNW	→	CF	112.271	0.000
TO	→	CF	75.349	0.000
UB	→	CF	83.233	0.000
GL	→	CF	52.757	0.000

for panel Fisher causality tests and

Table 12. Results of Hatemi-J causality tests.

Causality			Panel Fisher Stat.	Asymptotic Prob.	Causality			Panel Fisher Stat.	Asymptotic Prob.
FD+	→	EF+	86.570	0.000	FD+	→	CF+	168.249	0.000
FD+	→	EF−	81.270	0.000	FD+	→	CF−	83.442	0.000
FD−	→	EF−	63.022	0.000	FD−	→	CF−	110.736	0.000
FD−	→	EF+	39.905	0.011	FD−	→	CF+	64.087	0.000
NR+	→	EF+	37.609	0.020	NR+	→	CF+	280.565	0.000
NR+	→	EF−	62.587	0.000	NR+	→	CF−	48.677	0.001
NR−	→	EF−	47.886	0.001	NR−	→	CF−	106.533	0.000
NR−	→	EF+	150.303	0.000	NR−	→	CF+	94.083	0.000
HD+	→	EF+	81.905	0.000	HD+	→	CF+	65.271	0.000
HD+	→	EF−	42.149	0.003	HD+	→	CF−	50.499	0.000
HD−	→	EF−	14.059	0.827	HD−	→	CF−	251.103	0.000
HD−	→	EF+	176.458	0.000	HD−	→	CF+	130.227	0.000
TEC+	→	EF+	84.038	0.000	TEC+	→	CF+	92.336	0.000
TEC+	→	EF−	109.041	0.000	TEC+	→	CF−	54.675	0.000
TEC−	→	EF−	78.086	0.000	TEC−	→	CF−	68.994	0.000
TEC−	→	EF+	29.169	0.140	TEC−	→	CF+	18.080	0.701
RNW+	→	EF+	59.884	0.000	RNW+	→	CF+	86.185	0.000
RNW+	→	EF−	95.369	0.000	RNW+	→	CF−	60.048	0.000
RNW−	→	EF−	50.736	0.000	RNW−	→	CF−	62.146	0.000
RNW−	→	EF+	21.164	0.388	RNW−	→	CF+	22.542	0.312
TO+	→	EF+	45.632	0.002	TO+	→	CF+	46.283	0.000
TO+	→	EF−	142.617	0.000	TO+	→	CF−	77.094	0.000
TO−	→	EF−	34.459	0.044	TO−	→	CF−	176.040	0.000
TO−	→	EF+	90.522	0.000	TO−	→	CF+	66.133	0.000
UB+	→	EF+	57.493	0.000	UB+	→	CF+	91.620	0.000
UB+	→	EF−	96.569	0.000	UB+	→	CF−	80.180	0.000
UB−	→	EF−	149.198	0.000	UB−	→	CF−	146.812	0.000
UB−	→	EF+	47.123	0.001	UB−	→	CF+	190.579	0.000
GL+	→	EF+	140.591	0.000	GL+	→	CF+	144.062	0.000
GL+	→	EF−	67.021	0.000	GL+	→	CF−	79.336	0.000
GL−	→	EF−	20.905	0.527	GL−	→	CF−	71.274	0.000
GL−	→	EF+	38.905	0.014	GL−	→	CF+	11.476	0.967

for Hatemi-J causality tests. The panel Fisher causality results for our environmental models show that the main variables of FD, NR, HD, and TIC cause EF and CF. FD causes EF and CF, indicating the negative impact of FD on pollution. The Hatemi-J test also supports the inverted U-shaped relationship, with positive shocks in FD leading to positive causality between EF and CF and negative shocks in FD leading to negative causality between EF and CF. This finding suggests that the environmental impact of FD changes gradually. Positive shocks to FD increase the EF and CF, indicating that the early stages of FD increase environmental pressure due to increased energy consumption, industrial activities, and infrastructure investment. At this stage, the growth of the financial sector stimulates production and consumption activities with more investment. In this context, FD in the G-10 countries accelerates environmental degradation by increasing carbon emissions and natural resource consumption. In the later stage, negative shocks to FD reduce EF and CF, suggesting that environmental quality and sustainability are strengthened when financial development reaches a certain level in these countries. At this point, the financial system is more oriented towards clean energy technologies, green investments, and sustainable production methods, and investors and firms in these countries are beginning to allocate resources to switch to renewable energy and comply with environmental regulations.

NR causes EF and CF. This finding suggests that an increase in NR reduces environmental quality. The significance of positive and negative shocks in the Hatemi-J test also supports the negative impact of NR on pollution. NR in the G-10 countries is primarily used to increase the volume of production in environmentally damaging sectors. Moreover, these countries’ inadequate sustainable natural resource management may further deepen environmental degradation.

HD causes EF and CF, and this causality supports the finding that human development positively impacts environmental sustainability by reducing environmental pollution. High levels of education in the G-10 countries increase skilled human capital and individuals’ protective awareness of the environment, reducing the harmful effects of human activities on the environment in the long run. The causality of the positive and negative decompositions of HD on environmental pollution is also evidence of the impact of the HD factor on environmental quality in the G-10 countries.

The panel Fisher causality test shows that an increase in TIC leads to an increase in EF and CF. The advanced financial system and high economic growth of the G-10 countries encourage investments in technology and innovation. Technology development in these countries expands production volume, leading to more energy consumption and natural resource utilization. This has a negative effect on the environment, proving the rebound effect of TIC for G-10 countries. In Hatemi-J tests, positive shocks cause EF and CF, reinforcing technological innovation’s negative impact on environmental pollution.

Furthermore, RNW, TO, UB, and GL are the causes of the EF and CF of the G-10 countries. This strengthens the findings that RNW, TO, and GL factors reduce environmental pollution while supporting the negative effect of urbanization on environmental quality. Hatemi-J tests also find causality between positive and negative shocks in control variables and environmental pollution, supporting the long-run effect of these factors on EF and CF. Based on the causality findings, policymakers in the G-10 countries can formulate sound policies by considering these factors to ensure environmental sustainability.

5. Conclusions and Policy Recommendations

This study examined the effects of financial development, natural resource rent, technological innovation, and human development on ecological and carbon footprints for G-10 countries between 1990 and 2022. In addition, this study considered renewable energy, globalization, trade openness, and urbanization to be control variables affecting the ecological and carbon footprint. Kao cointegration and Westerlund’s DH cointegration tests, FMOLS and DOLS estimators, M and S robustness estimators, Panel Fisher causality tests, and Hatemi-J asymmetric causality tests were used to measure these variables’ long-term relationships and elasticities. Kao and DH cointegration tests confirm a long-run cointegration relationship. FMOLS and DOLS results reveal an inverted U-shaped link between financial development and environmental degradation. Furthermore, natural resource rent, technological innovation, and urbanization increase environmental degradation. Conversely, human development, renewable energy, globalization, and trade openness reduce environmental degradation. The results of Fisher’s test indicate a causal relationship between financial development, natural resource rent, human development, and technological innovation, and ecological and carbon footprints. According to the results of Hatemi-J tests, the negative shocks in ecological and carbon footprints are caused by the negative and positive shocks in financial development, natural resource rent, and technological innovation. Also, the positive and negative shocks in human development cause negative shocks in the carbon footprint, while the positive shocks in human development cause negative shocks in the ecological footprint. The negative and positive shocks in financial development, natural resource rent, and human development cause positive ecological and carbon footprint shocks. In contrast, the positive shock in technological innovation causes positive ecological and carbon footprint shocks. This study has some policy implications for the G-10 countries based on the findings.

(1) An inverted U-shaped link was found between financial development and environmental degradation. This implies that initially, development worsens environmental quality but improves it after a certain level. Therefore, it can be said that in these countries, financial flows expand industrial production in the initial stages and support green investments at a certain level of development. G-10 countries are composed of countries with strong financial structures. However, country-specific factors are important in the direction of this relationship. It is known that Sweden, Switzerland, the Netherlands, Belgium, the UK, and Germany support green investments. Green energy investments have started to increase in the US, and the green bond market has expanded in Canada. Therefore, these countries need to integrate their financial development processes with green policies and ensure that funds are channeled to green technology and energy, considering all these country-specific factors.

(2) The findings show that natural resource rent has a detrimental impact on environmental quality in G-10 countries and that these countries lack sustainable resource management. This phenomenon may be particularly pronounced in countries with large natural areas, such as the US and Canada, where natural resource rent can potentially exacerbate environmental degradation. The repercussions of the industrialization process are known to be ongoing in Germany. Therefore, these countries must implement sustainable resource management practices that consider country-specific factors, raise environmental taxes, and reduce energy-intensive resource consumption through sustainable investments.

(3) In G-10 countries, technological innovation increases environmental degradation. As a result, it can be said that economic growth in these countries is shaped by high technology-based production. These countries must integrate technological innovation with environmental policies. On the other hand, the US and Japan are globally important countries in technological innovation. However, the use of fossil fuels and industrial activities can put pressure on the environment. Therefore, developing country-specific innovation and environmental policies is important, especially considering countries with a high energy dependency.

(4) Human capital plays a vital role in promoting sustainable development. Training human capital can improve environmental quality and sustainable development outcomes. Integrating these insights into policy frameworks could lead to more sustainable practices in G-10 countries that benefit both the environment and economic growth.

(5) In G-10 countries, globalization and trade openness play an important role in improving environmental conditions in these countries. However, urbanization in these countries has a negative impact on environmental quality. Therefore, urban sustainability policies need to be prioritized. Various elements of urbanization, such as population growth, housing, industrialization, transport demand, and trade intensity, require meticulous planning. On the other hand, it is important to assess this relationship on a country-specific basis and draw policy implications. For example, the UK and Italy are trying to limit the negative impacts of urbanization through sustainable transportation systems, France is doing so by using nuclear energy, and Japan, Belgium, the Netherlands, Sweden, and Switzerland are doing so through green infrastructure and carbon-neutral city projects. Industrial production in large metropolitan areas in the US and the intensive use of natural resources in Canada cause high emissions. In this context, it is important for environmental sustainability that G-10 countries promote renewable energy, turn to green technologies, and invest in low-carbon city projects, considering country-specific factors.

The study has several limitations. First, the study considers financial development, natural resource rent, human development, technological innovation, globalization, trade openness, urbanization, and renewable energy, all affecting environmental degradation. It is important to note that these variables alone do not determine ecological and carbon footprints. Another limitation is that the analysis period of the study covers the period 1990–2022. Future researchers should consider improving the empirical study by integrating economic growth, economic complexity, economic uncertainty, governance factors, corruption, and stringent environmental policies into the model. Moreover, country-specific analysis can extend policy implications, and testing the model across different country groups can further strengthen the debate.