Revisiting Emissions: How Economic Structure, Financial Development, Urbanisation, Trade Openness, and Natural Resource Rent Shape CO₂ and N₂O

Abstract

Achieving zero carbon emissions is crucial for mitigating climate change and meeting global targets. This study examines the economic and financial drivers of carbon dioxide (CO₂) and nitrous oxide (N₂O) emissions using a panel dataset of 141 developed and developing countries from 1990 to 2020. Employing the generalised method of moments (GMM), the findings indicate that industrial and manufactural activities remain the dominant source of CO₂ emissions, particularly in developed economies, while agriculture is a major contributor to N₂O emissions, especially in developing countries. While the service sector reduces both emissions, the effect is more pronounced for CO₂ than for N₂O. Urbanisation, trade openness, and natural resource rents also positively correlate with emissions. However, financial development presents a dual effect, offering the potential for emissions reduction through green financing. These insights underscore the need for targeted policies, including stricter industrial regulations, sustainable agricultural practices, green urban planning, and financial strategies that support low-carbon transitions.

1. Introduction

Achieving zero carbon emissions remains a primary method for achieving the UN’s sustainable development goals (SDGs), particularly SDG 13 (Climate Action), SDG 7 (Affordable and Clean Energy), and SDG 11 (Sustainable Cities and Communities). Carbon dioxide (CO₂) and nitrous oxide (N₂O) are among the most critical greenhouse gases (GHGs), emitted through economic activities in both industrial and agricultural sectors and contributing significantly to global warming [1,2,3,4,5]. In 2024, global CO₂ emissions reached a historic high of 41.6 billion metric tons, up from 40.6 billion tons in 2023; meanwhile, N₂O—which has a global warming potential approximately 300 times that of CO₂—continues to rise (see Figure 1). These trends highlight the urgent need to investigate the structural and economic drivers of emissions in order to meet the Paris Agreement’s 1.5 °C target by 2050.

Theoretically, the Stochastic Impacts by Regression on Population, Affluence, and Technology (STIRPAT) framework provides a theoretical perspective for analysing the relationship between economic factors and environmental emissions, illustrating how economic growth, industrialisation, social structure, and natural resources influence environmental pressures [6]—an area in which previous studies have produced conflicting findings and left some theoretical gaps unaddressed. Although widely applied, the environmental Kuznets curve (EKC) hypothesis has yielded inconsistent results across pollutants and contexts. Indeed, some studies reported that economic growth reduces CO₂ emissions through increased emissions efficiency [7], while others found the opposite effect, especially in the long term due to industrialisation and high energy consumption [8,9]. While agriculture is recognised as a major source of N₂O emissions, its environmental impact remains controversial due to differing views on the role of agricultural trade in promoting sustainability versus increasing pollution [10,11]. Even the service sector (traditionally considered low-emission) has recently been found to generate significant indirect emissions through digital infrastructure and logistics [2,12]. Furthermore, financial development, trade openness, urbanisation, and resource rents all exhibit dual impacts on the environment—either promoting sustainability through green investment and low-carbon technologies, or exacerbating emissions in the context of “pollution havens” [13,14]. Despite these complexities, much research has relied on static models that overlook endogeneity and generalisability. For example, there is a lack of comprehensive frameworks integrating both CO₂ and N₂O emissions within a sectorally disaggregated and cross-national comparative analysis. Moreover, previous research has extensively analysed CO₂ emissions—highlighting industrialisation, energy consumption, urbanisation, and trade as significant contributors—but has relied on narrow proxies of economic performance (e.g., GDP growth or income per capita) while overlooking the heterogeneous roles of different economic sectors [15,16,17]. This gap limits the capacity to design targeted and sector-specific mitigation strategies.

This study makes three key contributions. First, it brings balanced attention to both CO₂ and N₂O emissions, the latter of which remains underexplored despite its substantial warming potential. Second, it advances methodological rigour by employing the generalised method of moments (GMM), which overcomes common issues of endogeneity and omitted variable bias in dynamic panel settings [18]. Third, and most notably, it builds on the STIRPAT framework—an extension of the IPAT model—to flexibly examine the influence of socio-economic and sectoral drivers on emissions [6]. In contrast to prior research, this study integrates a wide array of sector-specific variables, including agriculture (A), industry (I), manufacturing (M), services (S), financial development (F), urbanisation (P), trade openness (O), and natural resource rent (R) to disaggregate their respective contributions to environmental outcomes (see Figure 2). Moreover, this study’s novel contribution lies in its comparative analysis across developed and developing countries, shedding light on sectoral performance and environmental responsibility. These results provide new insights into achieving net zero carbon emissions, highlighting that emissions responsibilities vary by income level, sectoral composition, and technological capabilities.

The remainder of this study is structured as follows: Section 2 presents the theoretical framework, Section 3 describes the data and econometric methodology, Section 4 reports and discusses the empirical findings, and Section 5 concludes with policy implications.

2. Literature Review

Understanding what determines environmental impact is crucial for developing effective sustainability policies. According to Ehrlich and Holdren’s [19] IPAT model, population, consumption, and technology components are the main causes of environmental impact. The STIRPAT model, introduced by Dietz and Rosa [6], provides a flexible and empirically testable framework for assessing environmental pressures at various scales based on the IPAT model. They suggested the following formula of STIRPAT = I = aP^bA^cT^de, where “a”, “b”, and “c” are parameters, and “e” is the term of error. Since then, variations of the STIRPAT model have been proposed with different measurement variables assigned to I, P, A, and T. In this study, we extend the factors P, A, and T with different proxies, including the level of urbanisation representing the population, sectoral economic factors representing affluence, and the levels of financial development, resource exploitation, and integration representing a country’s level of technological adoption (Figure 2). By capturing the original spirit of the STIRPAT model, its extended version allows the components that impact the environment, and their directions, to be more precisely captured. The following section further discusses the mechanisms by which these components impact the environment. Overall, different sectors exhibit varied contributions to emissions, highlighting the need for sector-specific mitigation strategies. The differences in previous studies reflect the influence of production regions and study context on environmental outcomes. Thus, we fill this gap by conducting a detailed analysis of how production regions influence emissions across a diverse sample of countries, aiming to generalise the results and advance discussions on the attribution of sectoral responsibility.

2.1. Economic Growth and Emissions

The EKC hypothesis proposes that environmental degradation tends to intensify with economic growth in the early stages of development, but subsequently improves once a specific income level is reached [20]. However, prior research has yielded mixed results of this hypothesis in different economic contexts, thus challenging its generalisability. For example, Yusuf, Abubakar, and Mamman [17] applied an autoregressive distributed lag (ARDL) regression model to panel data from African OPEC countries between 1970 and 2016, finding that economic growth substantially impacted CO₂ and methane (CH₄) emissions, whereas its effect on N₂O emissions was insignificant. In the short term, they found that economic expansion was positively correlated with CH₄ emissions, but showed no statistically significant relationship with those of CO₂ and N₂O. Conversely, research by Niyonzima, Yao, and Ofori [7] suggested a short-term inverse relationship, where higher GDP levels coincide with reduced CO₂ emissions. However, Torun, Akdeniz, Demireli, and Grima [8] observed that a positive correlation often emerges in the long run, as economic growth leads to increased industrial activity and energy consumption, contributing to higher emissions. This cyclical relationship can be further intensified during periods of rapid economic expansion, which can potentially lead to the uncontrolled growth of emissions [9]. However, the failure to account for sectoral structure creates uncertainty about whether emissions reductions result from economic growth or merely shifts in sectoral composition. Moreover, the lack of research on sectoral differences in CO₂ and N₂O emissions represents a critical gap that needs to be addressed to refine our understanding of the EKC hypothesis. The next sections examine the mechanisms through which different economic activities affect emissions, providing the foundations for designing effective assessments that align economic development with sustainable environmental outcomes.

2.2. Productive Sectors and Emissions

The relationship between productive sectors and environmental degradation remains a critical research area. Agriculture, services, manufacturing, and industry significantly contribute to CO₂ and N₂O emissions, which are major GHGs driving global warming. CO₂ constitutes 82% of global emissions, while N₂O is 300 times more potent in warming potential and persists in the atmosphere for over a century [5]. These emissions stem from various economic activities, including fossil fuel combustion, synthetic fertiliser use, and industrial processes.

First, while the agricultural sector is essential for economic stability and food security, it is also a major contributor to GHG emissions. N₂O from synthetic fertilisers, livestock activities, and soil management account for over 40% of agricultural emissions [1,21]. Additionally, rice cropping systems contribute up to 19% of global methane emissions and 11% of agricultural N₂O emissions [4]. However, prior research has suggested that the link between agricultural activities and environmental sustainability is complex. For example, Balogh and Jámbor [22] highlighted the role of trade in exacerbating deforestation, transboundary pollution, and biodiversity loss. Meanwhile, Ghimire, Lin, and Zhuang [10] found that higher agricultural exports correlate with increased fertiliser and pesticide consumption, raising concerns about agricultural trade policies promoting environmental degradation. Conversely, agriculture can also absorb some emissions and use them as biomass, increasing crop yields through photosynthesis [23]. By shifting production to new locations and driving changes in technology and input use, agricultural trade can substantially enhance environmental sustainability [11].

Second, manufacturing and industrial production are vital for economic growth, job creation, and poverty alleviation [24,25]. However, the sector’s reliance on fossil fuels makes it a significant contributor to CO₂ and N₂O emissions [3]. On the one hand, Yusuf, Abubakar, and Mamman [17] reported that fossil fuel consumption in African OPEC nations had minimal long-term effects on N₂O emissions but significantly increased CO₂ emissions. Xie, Lu, and Xie [15] analysed panel data from China’s industrial sector from 2009 to 2021, revealing strong spatial correlations between industrial activity and carbon emissions. On the other hand, industrial activities have been found to follow the EKC hypothesis, where emissions initially rose with economic growth before declining at higher income levels in [26]. Moreover, Yu and Liu [27] found that changes in industrial structure have a significantly larger impact in the reduction of nitrogen oxide (NOx) emissions than economic growth in China. Therefore, the impact of industry on emissions remains controversial. Furthermore, manufacturing is the second-largest source of emissions after the energy sector, including electricity generation, gas production, steelmaking, and cement manufacturing [28,29]. Along with industrial production, manufacturing activities also contribute to increasing emissions, as mentioned by Roussilhe, Pirson, Xhonneux, and Bol [28], Rosita et al. [30], and Böckin and Tillman [31]. However, some studies have presented more complicated findings. For instance, Fujii and Managi [32] found a U-shaped EKC for CO₂ and an N-shaped relationship for N₂O, indicating sector-specific dynamics. These complexities emphasise the need to address emissions across different regions and industries to develop targeted mitigation strategies.

Third, fossil fuel dependence in transportation, consumption, and energy-intensive operations shapes the service sector’s environmental impact. While generally considered less emission-intensive than manufacturing, certain segments, including transportation, storage, and telecommunications, contribute significant direct emissions [33]. However, the service sector is responsible for 17–24% of national GHG emissions, contradicting its perceived low-emission status [2]. E-commerce and digital services have significantly increased energy consumption in data centres and logistics [12]. Additionally, rising consumer demand for travel and leisure has amplified emissions, particularly in urban areas with growing hospitality and transport industries [34]. Indirect emissions from supply chains and energy consumption substantially exacerbate the sector’s environmental footprint [35,36,37]. However, Hashmi et al. [38] found an inverted U-shaped relationship between the service sector and CO₂ emissions in Pakistan during 1971–2014 using a time series methodology, implying that it remains a low-polluting sector. It is noteworthy that, while many studies have addressed the relationship between the service sector and CO₂ emissions, few have done so with N₂O emissions. Furthermore, several mechanisms can determine the relationship between the service sector and N₂O emissions, such as emissions as a by-product of the operation of transport trucks [39], laundry and cleaning activities in hotels [40], and N₂O used in medical examination and treatment [41]. While previous studies have approached specific service activities and emissions separately, the lack of sector-level studies poses challenges in designing emission control policies tailored to the sector’s characteristics.

2.3. Financial Development and Emissions

In the literature, financial development has been shown to influence pollution through three interconnected pillars. First, financial growth stimulates economic expansion, increasing energy demand and CO₂ emissions, especially during positive financial shocks, such as fiscal easing policies and enhanced credit availability [42,43]. Second, financial development can drive investment into highly profitable yet environmentally harmful industries, particularly in developing nations with weak environmental regulations, where outdated technologies exacerbate pollution [44,45]. Finally, financial development is pivotal in fostering green technology adoption, with such mechanisms as green bonds and environmental investment funds enabling cleaner production [13,46].

While theoretical foundations provide different explanations for the effects of financial development on pollution, empirical studies have also yielded conflicting findings on this relationship. On the one hand, financial development supports environmental sustainability by financing clean energy transitions and reducing fossil fuel dependence [47,48]. Financial development also encourages green financial instruments, such as green bonds, which help fund environmentally beneficial projects [46,49]. Similarly, Sheraz et al. [50] found that, in G20 countries, green credit mechanisms and low-carbon projects led to a decline in CO₂ emissions from 1986 to 2018. By directing resources towards clean energy, financial development can support sustainable economic growth while mitigating environmental degradation. In Indonesia and Malaysia, Shahbaz et al. [51,52] observed that financial development stages were accompanied by lower CO₂ emissions, as a well-functioning financial sector encourages investors to adopt green technology. On the other hand, financial development can worsen air pollution when it primarily supports energy consumption and energy-intensive industries [53,54]. Tran [55] indicated that the impacts of financial development on CO₂ emissions depend on economic growth and energy consumption. In economies reliant on natural resources, increased financial development often results in the expansion of high-emission activities [56]. For example, Jiang and Ma [57] found that financial development increased CO₂ emissions across 155 countries, while Ma and Fu [53] confirmed a similar trend in 120 nations from 1991 to 2014, particularly in developing economies. However, Nguyen et al. [58] confirmed a U-shaped relationship between financial development and pollutants in the long term in Vietnam, claiming that the financial system increases emissions during its development. These contrasted findings highlight the importance of considering financial development as a key variable in economic-environmental studies.

2.4. Openness and Emissions

Some studies have proposed that trade openness facilitates efficiency in resource allocation, supports technology transfer, and enhances productivity, contributing to sustainable development [59,60]. However, trade openness has been linked to both positive and negative environmental outcomes. For example, Salman et al. [61] noted that exports, particularly in certain Asian economies, contribute more to CO₂ emissions than imports. This suggests that trade-driven economic growth may intensify emissions, particularly in nations where production relies on fossil fuels. Moreover, an economy’s industrial composition influences how trade affects emissions. Shen, Liu, and Tian [14] found that trade significantly increased logistics-related carbon emissions, especially in such sectors as textiles, sewing, and leather product manufacturing, in the chemical industry, metal product manufacturing, and machinery and equipment manufacturing in China’s economic belt from 2017 to 2021. Rehan, Gungor, Qamar, and Naz [16] further highlighted the positive role of international trade in CO₂ emissions caused by the import and export of goods with higher carbon footprints in G20 and EU countries. However, other studies have challenged the direct link between trade and emissions. Sharma [62] found that trade openness has negative effects on CO₂ emissions, using data from 69 countries between 1985 and 2005. Similarly, Le et al. [63] found that trade helps reduce pollution in high-income countries, but worsens it in middle- and low-income nations, reflecting differences in regulatory standards and industrial structures. However, Nguyen et al. [64] noted that global trade is more environmentally beneficial to countries facing environmental challenges compared to those with higher environmental standards. Since trade depends on production and transportation, this distinction underscores the complexity of assessing trade openness across different contexts. Given these diverse perspectives, this study examines trade within a global framework to provide a broader understanding of its environmental impact.

2.5. Urbanization and Emissions

Urbanisation has significantly contributed to rising emissions, straining natural resources, and disrupting ecological systems [65]. Increased energy demand due to global urbanisation has led to higher carbon emissions and intensified water consumption [66], revealing complex interdependencies between urban growth and environmental sustainability. While sustainable development frameworks seek to balance present needs with future sustainability, urban expansion often exacerbates environmental degradation through excessive resource use, pollution, and water-use inefficiencies [65]. However, previous studies have shown that urbanisation has a complex relationship with emissions, as it can both mitigate and exacerbate environmental impacts. Indeed, high population densities in cities allow for more efficient public transportation systems, reducing per capita emissions compared to rural areas reliant on personal vehicles [67]. Moreover, urbanisation can drive innovation in clean technologies and sustainable practices, as seen in cities adopting electric mobility, renewable energy integration, and climate-resilient policies [68,69]. However, city expansion often leads to increased energy consumption, congestion, and higher transportation emissions. For example, the construction and operation of buildings contribute significantly to CO₂ emissions due to reliance on cement, steel, and fossil fuel-based energy sources [70,71]. Traffic congestion in densely populated cities exacerbates air pollution, increasing GHG emissions and such harmful pollutants as NOx and particulate matter (PM) [72,73]. Furthermore, as urban populations grow, the demand for energy-intensive goods and services rises, reinforcing the link between urbanisation and higher emissions [14,37]. These contrasting findings highlight urbanisation as a critical factor in shaping emissions within broader socio-economic dynamics.

2.6. Natural Resource Rents and Emissions

Natural resource rents (i.e., the revenues derived from the extraction of natural resources) play a significant role in pushing economic development, but have a complex relation to environmental outcomes [74]. On the positive side, resource-rich economies can reduce their dependence on fuel imports and switch to advanced technologies that incorporate recycling, thereby reducing environmental degradation [75]. Natural resource rents can also be allocated to intergenerational wealth transfers and investments in human, produced, and natural capital, along with stabilisation measures against economic volatility [76]. This enhances economic resilience and supports progress towards sustainable development. For example, Tufail et al. [77] revealed that natural resource rent improved the natural environment by reducing carbon emissions for OECD economies during 1990–2018. Similarly, Musah et al. [78] gave empirical evidence that resource rents reduce ecological footprints in a sample of West African states. Moreover, an abundance of natural resources can attract foreign direct investments in eco-friendly technologies, further alleviating environmental pressures [79]. Indeed, examining resource rents across 90 Belt and Road Initiative economies, Zuo et al. [80] also found that resource rents, coupled with technological innovation, reduce ecological footprints. These studies confirmed that higher natural rents could enable a successful transition towards sustainability by reinvesting resource revenues into green initiatives.

However, in many cases, high reliance on natural rents is associated with increased emissions due to the “resource curse” or “carbon lock-in” [81,82]. Economies dependent on resource extraction often prioritise short-term economic gains over long-term environmental sustainability, leading to excessive fossil fuel consumption, deforestation, and pollution [83,84]. The expansion of extractive industries increases CO₂ and CH₄ emissions, particularly in oil- and coal-dependent economies where industrial activities are heavily reliant on fossil fuels [84,85]. Additionally, poor institutional quality and the inefficient allocation of resource revenues can prevent investments in green technologies, leading to environmental degradation and prolonged dependence on carbon-intensive industries [86,87]. This dependency reinforces a cycle where natural rents drive economic growth and contribute significantly to emissions, making it challenging for such economies to achieve sustainable environmental outcomes. Due to these contradictions, this study recommends a more careful approach to natural resource revenue in a broader context to reflect its role in sustainable development.

3. Data and Methodology

3.1. Data

The data were primarily sourced from the World Bank’s World Development Indicators (WDI), including emission indicators (CO₂ and N₂O), economic growth, natural resource rents, trade openness, and urbanisation. Financial development data were sourced from the International Monetary Fund (IMF) based on Svirydzenka [88]. This study covers 141 developed and developing countries from 1990 to 2020, yielding an unbalanced panel dataset with 4135 observations.

3.2. Model

This study extends Dietz and Rosa’s [6] STIRPAT model, based on Ehrlich and Holdren’s [19], to assess the relationship between environment and socio-economic-technological factors (see Figure 2). The adapted model, incorporating additional relevant factors, is presented as follows

$${EMS}_{it}=\alpha +\gamma \ast {EGR}_{it}+{\sum }_{i=1}^4{\theta }_i\ast {Control}_i+{\epsilon }_{it}$$

(1)

where EMS is measured emission using the natural logarithm of CO₂ and N₂O emissions. As recent studies have suggested that air pollution may persist over time [53,89], our model includes a lagged emission variable, leading to the revised equation:

$$\begin{gathered} {EMS}_{it}={\alpha }_{\mathrm{a}}+{\beta }_a\ast {EMS}_{it-1}+{\gamma }_a\ast {EGR}_{it}+{\theta }_{1\mathrm{a}}\ast {NRR}_{it}+{\theta }_{2\mathrm{a}}\ast {TRO}_{it}+{\theta }_{3\mathrm{a}} \\ \ast {URB}_{it}+{\theta }_{4\mathrm{a}}\ast {FID}_{it}+{\epsilon }_{it} \end{gathered}$$

(2)

Then, we proceed to disaggregate economic activities into the components of sectoral activities (SEC), and extend Equation (2) as follows

$$\begin{gathered} {EMS}_{it}={\alpha }_{\mathrm{b}}+{\beta }_b\ast {EMS}_{it-1}+{\gamma }_b\ast {SEC\left(AGR\right)}_{it}+{\theta }_{1\mathrm{b}}\ast {NRR}_{it}+{\theta }_{2\mathrm{b}}\ast {TRO}_{it} \\ +{\theta }_{3\mathrm{b}}\ast {URB}_{it}+{\theta }_{4\mathrm{b}}\ast {FID}_{it}+{\epsilon }_{it} \end{gathered}$$

(3)

$$\begin{gathered} {EMS}_{it}={\alpha }_{\mathrm{c}}+{\beta }_c\ast {EMS}_{it-1}+{\gamma }_c\ast {SEC\left(IND\right)}_{it}+{\theta }_{1\mathrm{c}}\ast {NRR}_{it}+{\theta }_{2\mathrm{c}}\ast {TRO}_{it} \\ +{\theta }_{3\mathrm{c}}\ast {URB}_{it}+{\theta }_{4\mathrm{c}}\ast {FID}_{it}+{\epsilon }_{it} \end{gathered}$$

(4)

$$\begin{gathered} {EMS}_{it}={\alpha }_{\mathrm{d}}+{\beta }_d\ast {EMS}_{it-1}+{\gamma }_d\ast {SEC\left(MAF\right)}_{it}+{\theta }_{1\mathrm{d}}\ast {NRR}_{it}+{\theta }_{2\mathrm{d}}\ast {TRO}_{it} \\ +{\theta }_{3\mathrm{d}}\ast {URB}_{it}+{\theta }_{4\mathrm{d}}\ast {FID}_{it}+{\epsilon }_{it} \end{gathered}$$

(5)

$$\begin{gathered} {EMS}_{it}={\alpha }_{\mathrm{e}}+{\beta }_e\ast {EMS}_{it-1}+{\gamma }_e\ast {SEC\left(SER\right)}_{it}+{\theta }_{1\mathrm{e}}\ast {NRR}_{it}+{\theta }_{2\mathrm{e}}\ast {TRO}_{it} \\ +{\theta }_{3\mathrm{e}}\ast {URB}_{it}+{\theta }_{4\mathrm{e}}\ast {FID}_{it}+{\epsilon }_{it} \end{gathered}$$

(6)

where AGR, IND, MAF, and SER represent agriculture, industry, manufacturing, and services, respectively. All other definitions of variables are presented in

Table 1. Definitions of variables.

Variable	Symbol	Measurement	Sources
Carbon emission	CO2	Natural logarithm of CO2 emissions (kt)	WDI
Nitrous oxide emissions	N2O	Natural logarithm of N2O emissions (thousand metric tons of CO2 equivalent)	WDI
Economic growth	EGR	Natural logarithm of GDP per capita	WDI
Agriculture	AGR	The ratio of value added in agriculture, forestry, and fishing to GDP	WDI
Industry	IND	The ratio of value added in industry (including construction) to GDP	WDI
Manufacturing	MAF	The ratio of value added in manufacturing to GDP	WDI
Services	SER	The ratio of value added in services to GDP	WDI
Natural resource rents	NRR	The ratio of total natural resources rents to GDP	WDI
Trade openness	TRO	The ratio of the sum of export and import of goods and services to GDP	WDI
Urbanisation	URB	Natural logarithm of urban population	WDI
Financial development	FID	Financial development index	IMF

.

3.3. Econometrical Method

Economic development requires promoting related activities, including industry, agriculture, manufacturing, and services. Increasing these can lead to the over-exploitation of natural resources, environmental degradation and, ultimately, a decline in human well-being [90]. In essence, while economic growth can bring immediate benefits, it also carries long-term environmental and societal consequences [91]. Moreover, climate change, pollution, and biodiversity loss may significantly impact the economy. These changes typically generate negative impacts, where the depletion of natural resources or the occurrence of environmental disasters can hinder productivity and escalate the costs associated with economic activities [92]. This interdependent relationship between economic development and environmental quality suggests the presence of a potential endogeneity problem.

Thus, panel regression in this case may face two key challenges: (a) the issue of reverse causality and (b) the presence of serial correlation due to the inclusion of lagged dependent variables [93]. To address these concerns, the system generalized method of moments (GMM) estimator is particularly effective, as it helps mitigate endogeneity stemming from both reverse causality and serial correlation. Traditional panel estimation methods, including OLS, FEM, and REM, often yield biased results in the presence of endogeneity in economic research. GMM, based on Hansen’s [94] framework, is particularly effective for short panel datasets (the number of individuals (N) is many times larger than the number of times (T)). It relies on moment conditions derived from observational data to estimate unknown parameters. Among the main GMM approaches, difference-GMM (diff-GMM) [95] eliminates unobserved effects by using first-differenced variables and their lags as instruments. However, this approach may lead to information loss, particularly for time-invariant variables. System-GMM [18] offers an improved method by transforming instrumental variables and incorporating additional tools to enhance the effectiveness of estimations by applying a preliminary estimation with an initial weight matrix and then optimising the moment conditions using a corrected variance matrix. Beck et al. [96] showed that, much like conventional two-stage least squares methods, system-GMM is capable of handling the complexities of reverse causality in dynamic panel models. However, excessive number of instrumental variables used in GMM can distort results. To mitigate this risk, Roodman [97] suggested limiting the number of instruments to the number of groups (N). The validity of instruments is verified through the Hansen–Sargan test (ensuring a p-value > 0.1), while the AR(2) test confirms the absence of second-order autocorrelation in residuals, reinforcing the reliability of the estimations.

4. Findings and Discussions

4.1. Findings

Table 2. Descriptive statistics.

Variable	Obs	Mean	Std. dev.	Min	Max
CO2	4135	9.625	2.458	3.178	16.213
N2O	4135	8.088	2.232	−0.476	13.221
EGR	4135	3.267	4.995	−50.248	88.958
AGR	4135	12.893	11.869	0.000	64.673
IND	4135	26.923	11.079	0.000	74.113
MAF	4135	12.798	7.236	0.000	49.879
SER	4135	51.484	13.213	0.000	87.421
NRR	4135	6.305	9.840	0.000	79.431
TRO	4135	80.936	48.793	9.955	437.327
URB	4135	15.249	1.975	10.018	20.584
FID	4135	0.309	0.233	0.000	1.000

presents descriptive statistics for the research variables. The average logarithm of CO₂ level is 9.62 (Std. dev. = 2.46), with Kiribati having the highest and China having the lowest emissions, corresponding to their CO₂ values. The average logarithm of N₂O level is 8.81 (Std. dev. = 2.23), with the Marshall Islands and China showing the highest and lowest emissions, respectively. Economic growth (EGR) has a mean value of 3.267 with a substantial standard deviation of 4.995, reflecting high variability in GDP growth across observations. The agricultural sector (AGR) accounts for an average of 12.893% of economic activity, although its high standard deviation of 11.869 suggests substantial cross-country differences, ranging from 0.000% to 64.673%. In contrast, the industrial sector (IND) has a higher average contribution of 26.923%, with values spanning from 0.000% to 74.113%, thus reinforcing its role as a dominant driver of emissions and economic output. Natural resource rents account for 6.31% of GDP on average. Average trade openness (OP) is 80.94%, indicating a high degree of integration of economies. According to the IMF, urbanisation averages 15.25% of the total population, while the financial development index remains at 0.309. These descriptive statistics highlight the diverse economic and environmental conditions across different economies and lay the foundation for subsequent analyses of emissions determinants.

Table 3. Correlation matrix.

	CO2	N2O	EGR	AGR	IND	MAF	SER	NRR	TRO	URB	FID
CO2	1
N2O	0.773 ***	1
EGR	−0.014	0.046 ***	1
AGR	−0.522 ***	−0.096 ***	0.052 ***	1
IND	0.383 ***	0.266 ***	0.094 ***	−0.327 ***	1
MAF	0.413 ***	0.326 ***	0.015	−0.330 ***	0.324 ***	1
SER	0.177 ***	−0.055 ***	−0.168 ***	−0.558 ***	−0.320 ***	0.055 ***	1
NRR	−0.050 ***	−0.021	0.104 ***	0.144 ***	0.500 ***	−0.160 ***	−0.476 ***	1
TRO	−0.099 ***	−0.333 ***	0.050 ***	−0.309 ***	0.076 ***	0.114 ***	0.177 ***	−0.055 ***	1
URB	0.867 ***	0.918 ***	0.028 *	−0.183 ***	0.288 ***	0.348 ***	0.016	−0.012	−0.332 ***	1
FID	0.620 ***	0.298 ***	−0.075 ***	−0.654 ***	0.028 *	0.238 ***	0.525 ***	−0.261 ***	0.223 ***	0.365 ***	1

Notes: *, *** show statistical significance at 10% and 1%, respectively.

presents correlation coefficients between key variables. EGR exhibits a strong positive correlation with the logarithm of N₂O, while its correlation with the logarithm of CO₂ is not statistically significant. There is also a moderate correlation between IND and the logarithm of CO₂, indicating the sector’s contribution to environmental degradation. AGR, in contrast, shows a weaker or even negative correlation with emissions, implying that agricultural economies may have a smaller carbon footprint. Meanwhile, TRO shows a negative correlation with emissions, and URB and FID display positive correlations with pollution. Furthermore, the correlation coefficients among the independent variables are below 0.8, indicating no high multicollinearity among the independent variables.

Table 4. The impact of economic activities on emission.

	EMS (CO2)	EMS (N2O)
L.EMS	0.9795 ***	0.9825 ***
	(889.20)	(766.58)
EGR	0.0105 ***	0.0031 ***
	(62.30)	(13.48)
NRR	0.0003 ***	0.0004 ***
	(5.18)	(3.83)
TRO	0.0001 **	0.0001 ***
	(2.44)	(3.63)
URB	0.0164 ***	0.0280 ***
	(9.13)	(14.47)
FID	0.0240 **	−0.0645 ***
	(2.39)	(−10.35)
Constant	−0.0852 ***	−0.2771 ***
	(−4.14)	(−13.24)
Obs	3994	3994
p-value of AR(2)	0.6316	0.1129
p-value of Hansen test	0.1502	0.4096
Number of groups	141	141
Number of IV	99	111

Notes: **, *** show statistical significance at 5%, and 1%, respectively. ( ) is z-value.

,

Table 5. The impact of sectoral production on CO₂.

	SEC (AGR)	SEC (IND)	SEC (MAF)	SEC (SER)
	EMS (CO2)	EMS (CO2)	EMS (CO2)	EMS (CO2)
L.EMS	0.9795 ***	0.9641 ***	0.9695 ***	0.9710 ***
	(229.37)	(822.85)	(1263.58)	(959.92)
SEC	0.0013 **	0.0018 ***	0.0004 *	−0.0014 ***
	(1.98)	(13.71)	(1.84)	(−11.90)
NRR	0.0001	−0.0002	0.0004 ***	0.0002 *
	(0.81)	(−1.57)	(4.07)	(1.89)
TRO	0.0001	0.0000 *	0.0001 ***	0.0000 **
	(1.55)	(1.68)	(2.85)	(2.49)
URB	0.0370 ***	0.0376 ***	0.0389 ***	0.0320 ***
	(9.73)	(29.07)	(30.73)	(29.07)
FID	−0.0238	0.0277 ***	−0.0034	0.0519 ***
	(−1.29)	(3.80)	(−0.55)	(6.17)
Constant	−0.3607 ***	−0.2635 ***	−0.2877 ***	−0.1337 ***
	(−9.90)	(−17.78)	(−18.83)	(−7.77)
Obs	3994	3994	3994	3994
p-value of AR(2)	0.9823	0.9959	0.9672	0.9951
p-value of Hansen test	0.1339	0.4196	0.2906	0.2355
Number of groups	141	141	141	141
Number of IV	110	133	132	131

Notes: *, **, *** show statistical significance at 10%, 5%, and 1%, respectively. ( ) is z-value.

and

Table 6. The impact of sectoral production on N₂O.

	SEC (AGR)	SEC (IND)	SEC (MAF)	SEC (SER)
	EMS (N2O)	EMS (N2O)	EMS (N2O)	EMS (N2O)
L. EMS	0.9876 ***	0.9919 ***	0.9567 ***	0.9833 ***
	(253.99)	(1737.09)	(132.24)	(310.67)
SEC	0.0024 ***	0.0002 ***	0.0012 ***	−0.0008 ***
	(5.04)	(3.42)	(2.82)	(−2.93)
NRR	0.0003 *	0.0002 ***	0.0011 ***	0.0007 **
	(1.77)	(3.79)	(3.19)	(2.48)
TRO	0.0001 *	0.0000 ***	0.0002 ***	0.0001 ***
	(1.69)	(4.16)	(3.48)	(4.60)
URB	0.0157 ***	0.0089 ***	0.0479 ***	0.0201 ***
	(3.27)	(8.84)	(6.18)	(6.34)
FID	0.0213	−0.0493 ***	−0.0797 ***	−0.0321 ***
	(1.29)	(−17.41)	(−5.05)	(−3.03)
Constant	−0.1735 ***	−0.0536 ***	−0.3853 ***	−0.1252 ***
	(−3.54)	(−4.71)	(−5.99)	(−4.25)
Obs	3994	3994	3994	3994
p-value of AR(2)	0.1203	0.1254	0.1307	0.1340
p-value of Hansen test	0.2632	0.3155	0.2708	0.5352
Number of groups	141	141	141	141
Number of IV	87	132	89	99

Notes: *, **, *** show statistical significance at 10%, 5%, and 1%, respectively. ( ) is z-value.

assess the validity of the GMM regression results through the AR(2) and Hansen tests, which examine autocorrelation and the validity of instrumental variables, respectively. The p-values for both tests are over 0.1, indicating that the null hypothesis is not rejected for both N₂O and CO₂. This suggests that the model errors are not autocorrelated after applying GMM estimation and that the instrumental variables used are appropriate, as they are uncorrelated with the model residuals [98].

The results in Table 4 highlight the significant and persistent influence of economic activities on CO₂ and N₂O emissions. The lagged emission variable (L.EMS) exhibits highly significant coefficients for CO₂ (0.9795) and N₂O (0.9825), with 1% statistical significance, indicating that past emissions strongly determine current levels. This persistence suggests that emissions tend to accumulate over time, reinforcing the long-term environmental impact of economic activities. Additionally, EGR has a positive and statistically significant effect on both CO₂ (0.0105) and N₂O (0.0031), implying that, as GDP per capita rises, emissions of both GHGs increase with a 1% statistical significance. However, the smaller coefficient for N₂O suggests N₂O’s relatively lower sensitivity to economic expansion compared to CO₂. These findings underscore the environmental trade-offs associated with economic growth. Moreover, NRR exhibits a small but statistically significant positive impact on both CO₂ (0.0003 with 10% statistical significance) and N₂O (0.0004 with 1% statistical significance) emissions, suggesting that economies reliant on resource rents tend to experience higher emissions. TRO also shows a significant but modest positive relationship with emissions (0.0001 for CO₂ and 0.0001 for N₂O) with statistical significance, indicating that increased trade activity is associated with higher emissions, likely due to the hypothesis of expansion of industrial production and transportation-related energy consumption.

Moreover, URB has a pronounced positive effect on emissions, with coefficients of 0.0164 * for CO₂ and 0.0280 for N₂O, suggesting that growing urban populations drive higher energy consumption and waste production, thereby increasing GHGs. FID presents mixed effects, with a significant positive association with CO₂ emissions (0.0240), suggesting that increased financial resources may facilitate investments in emission-intensive industries. However, FID has a strongly negative effect on N₂O emissions (−0.0645) with statistical significance, implying that the hypothesis of financial development could support technologies or policies that mitigate N₂O, such as improved agricultural practices.

To examine the sectoral impacts of economic activities, economic activities are categorised into AGR, IND, MAF, and SER, and the below findings are estimated from Equations (3)–(6). Table 5 and Table 6 present the results on the relationship between economic activities and CO₂/N₂O emissions.

The results from Table 5 and Table 6 reveal key similarities and differences in the impact of economic activities on CO₂ and N₂O emissions, underscoring the sector- and gas-specific dynamics of environmental degradation. The persistence of emissions is evident for both GHGs, as indicated by the significant coefficients of the lagged emissions variable, which range from 0.9641 to 0.9795 for CO₂ (Table 5) and from 0.9567 to 0.9919 for N₂O (Table 6) across different sectors, similar to the findings reported in Table 4. Although N₂O emissions exhibit slightly higher persistence levels, particularly in the industrial sector (0.9919) with 1% statistical significance, reflecting its long atmospheric lifespan and the sustained impact of industrial activities on N₂O release, sectoral production affects CO₂ and N₂O emissions in distinct ways. In the agriculture sector, AGR significantly increases emissions of both gases, but with a stronger effect for N₂O (0.0024, compared to 0.0013 for CO₂), likely due to the intensive use of nitrogen-based fertilisers that drive N₂O emissions. In contrast, IND contributes more to CO₂ (0.0018) than to N₂O emissions (0.0002), emphasising the predominance of fossil fuel combustion in industrial activities. MAF also exhibits a positive and significant impact on both gases, but its influence is relatively small for both CO₂ (0.0004) and N₂O (0.0012), suggesting that emissions depend on production processes and energy efficiency measures. Notably, SER reduces emissions of both gases, but with a more pronounced effect for CO₂ (−0.0014, compared to −0.0008 for N₂O), implying that transitioning towards service-based economies may be more effective in mitigating carbon emissions than N₂O.

URB consistently shows a significant and positive impact on emissions in all three tables, reinforcing the notion that increasing urbanisation drives higher emissions through intensified energy demand, industrialisation, and transportation expansion. However, the effect of URB is stronger for CO₂ emissions in Table 5 (ranging from 0.0320 to 0.0389) compared to its impact in Table 4 and Table 6, where the effect varies across sectors (from 0.0089 to 0.0479 for N₂O). They suggest that while urbanisation contributes significantly to CO₂ and N₂O emissions, its influence on the former is more pronounced, particularly in the participation of explanatory variables of industries in empirical models. FID presents mixed findings across the tables. In Table 6, FID has a negative and significant effect on N₂O emissions across all sectors (ranging from −0.0321 to −0.0797) except agriculture (insignificance), aligning with the results in Table 4. However, in Table 5, FID’s impact on CO₂ emissions varies by sector—it is significantly positive in the industrial (0.0277) and service (0.0519) sectors, but insignificant in agriculture and manufacturing. These findings suggest that financial development fuels carbon emissions primarily in industries that rely on capital-intensive production and energy consumption.

Table 7. Univariate comparison between developed and developing countries.

Variables	Developing Countries (A)	Developed Countries (B)	(C) = (A) − (B)	T Statistics
CO2	9.085	11.170	−2.086	−25.733 ***
N2O	7.881	8.681	−0.800	−10.216 ***
EGR	3.501	2.596	0.906	5.122 ***
AGR	16.470	2.645	13.825	38.139 ***
IND	27.477	25.336	2.141	5.462 ***
MAF	11.935	15.269	−3.333	−13.245 ***
SER	47.806	62.021	−14.215	−34.347 ***
NRR	8.103	1.154	6.949	20.912 ***
TRO	72.756	104.367	−31.610	−19.025 ***
URB	15.079	15.738	−0.659	−9.494 ***
FID	0.207	0.600	−0.393	−70.372 ***

Notes: *** show statistical significance at 1%, respectively.

presents a univariate analysis where the total sample is divided into developed and developing countries, following the classification recommended by UNCTAD (https://unctadstat.unctad.org, accessed on 12 November 2024). Column (A) reports the mean values of the variables for developing countries, while column (B) provides the corresponding means for developed countries. Given that the variables follow a normal or approximately normal distribution, a t-test is employed to compare the two groups, with the results displayed in column (C). The findings indicate that all variables have p-values below 0.01, suggesting statistically significant differences between the two groups. These results support the hypothesis that economic growth affects emissions differently in developed and developing economies.

To further investigate these disparities, an interaction variable between the dummy variable for country classification (developing and developed countries, according to UNCTAD) and economic activities was incorporated into the research model, followed by a re-estimation of Equations (3)–(6). The results are presented in

Table 8. The impact of interaction between country classification and sectoral production on CO₂.

	SEC (AGR)	SEC (IND)	SEC (MAF)	SEC (SER)
	EMS (CO2)	EMS (CO2)	EMS (CO2)	EMS (CO2)
L.EMS	0.9925 ***	0.9729 ***	0.9900 ***	0.9787 ***
	(253.23)	(379.92)	(309.18)	(367.19)
SEC	0.0041 ***	0.0020 ***	0.0032 ***	−0.0022 ***
	(6.68)	(6.62)	(6.27)	(−7.65)
SEC × DMY	0.0117 **	−0.0041 ***	−0.0045 ***	−0.0010 ***
	(2.24)	(−7.79)	(−3.83)	(−6.02)
NRR	0.0007 ***	0.0005 **	0.0028 ***	0.0015 ***
	(3.37)	(2.28)	(8.32)	(4.31)
TRO	0.0001 **	0.0001 **	−0.0000	0.0001 **
	(2.35)	(2.14)	(−0.59)	(2.58)
URB	0.0258 ***	0.0001	−0.0064	0.0102 **
	(5.03)	(0.04)	(−1.12)	(2.55)
FID	0.0066	0.2315 ***	0.1121 ***	0.1854 ***
	(0.31)	(9.44)	(4.53)	(8.32)
Constant	−0.3790 ***	0.1677 ***	0.1473 **	0.1267 **
	(−6.93)	(4.39)	(2.52)	(2.43)
Obs	3994	3994	3994	3994
p-value of AR(2)	0.9925	0.9571	0.8871	0.8743
p-value of Hansen test	0.1489	0.1175	0.1290	0.1589
Number of groups	141	141	141	141
Number of IV	105	124	93	113

Notes: **, *** show statistical significance at 5%, and 1%, respectively. ( ) is z-value.

and

Table 9. The impact of interaction between country classification and sectoral production on N₂O.

	SEC (AGR)	SEC (IND)	SEC (MAF)	SEC (SER)
	EMS (N2O)	EMS (N2O)	EMS (N2O)	EMS (N2O)
L. EMS	0.8988 ***	0.9662 ***	0.9886 ***	0.9713 ***
	(85.07)	(148.52)	(247.60)	(231.25)
SEC	0.0018 ***	0.0009 **	0.0004 **	−0.0014 ***
	(4.49)	(2.51)	(2.44)	(−5.98)
SECxDMY	0.0281 ***	−0.0018 **	−0.0036 ***	−0.0004 **
	(5.70)	(−2.33)	(−7.15)	(−1.97)
NRR	0.0012 ***	−0.0004	0.0002 *	0.0003
	(5.52)	(−1.25)	(1.76)	(0.92)
TRO	−0.0001	0.0000	0.0001 ***	0.0002 ***
	(−1.42)	(0.35)	(2.93)	(4.49)
URB	0.1042 ***	0.0350 ***	0.0126 ***	0.0331 ***
	(10.39)	(5.11)	(2.81)	(6.93)
FID	−0.0158	−0.0473	0.0092	−0.0346
	(−0.75)	(−1.44)	(0.84)	(−1.59)
Constant	−0.7979 ***	−0.2457 ***	−0.0944 ***	−0.1884 ***
	(−9.62)	(−4.39)	(−2.60)	(−4.97)
Obs	3994	3994	3994	3994
p−value of AR(2)	0.1385	0.1273	0.1231	0.1386
p−value of Hansen test	0.3126	0.3603	0.4325	0.1111
Number of groups	141	141	141	141
Number of IV	113	63	87	91

Notes: *, **, *** show statistical significance at 10%, 5%, and 1%, respectively. ( ) is z-value.

, respectively. DMY is a dummy variable, equalling 0 if the country is developing and 1 if the country is developed.

The findings from Table 8 and Table 9 provide deeper insights into how sectoral production affects CO₂ and N₂O emissions when considering country classifications, complementing the results from Table 5 and Table 6. A key finding is that the interaction term SEC*DMY exhibits significant variation across sectors, indicating that the relationship between economic activities and emissions is not uniform across different economies. For CO₂ emissions in Table 8, SEC*DMY is positive and significant in the AGR sector (0.0117), suggesting that agricultural production in developed countries leads to greater carbon emissions than in developing economies. Conversely, IND (−0.0041), MAF (−0.0045), and SER (−0.0010) sectors in developed countries contribute less to CO₂ emissions compared to developing countries, indicating that advanced economies may benefit from cleaner technologies, greater energy efficiency, and stronger environmental regulations in these sectors. These conclusions are similar to those for N₂O emissions in Table 9, where SEC*DMY is strongly positive in the agricultural sector (0.0281), indicating that developed countries produce significantly higher N₂O emissions from agricultural activities than developing nations. Meanwhile, IND (−0.0018), MAF (−0.0036), and SER (−0.0004) sectors in developed countries generate lower N₂O emissions than their developing counterparts.

4.2. Discussions

The above findings provide empirical insights into the relationship between economic activities and environmental degradation. Economic growth plays a significant role in shaping emissions, confirming the EKC hypothesis. The findings reveal that economic expansion is positively associated with emissions, particularly in the early stages of development, consistent with Yusuf, Abubakar, and Mamman [17]. However, the smaller coefficient for N₂O compared to CO₂ suggests that industrial and energy production sectors contribute more significantly to carbon emissions, whereas N₂O emissions primarily arise from agricultural activities. The findings also indicate that industrial activities significantly contribute to environmental degradation through high fossil fuel consumption, aligning with Xie, Lu, and Xie [15] and Roussilhe, Pirson, Xhonneux, and Bol [28], as well as with Bui, Van Nguyen, Huynh, Bui, Nguyen, Perng, Bui, and Nguyen [1]. Similarly, manufacturing also positively affects emissions, reinforcing the notion that energy-intensive production processes exacerbate environmental stress, as supported by Lee, McJeon, Yu, Liu, Kim, and Eom [29]. Conversely, the service sector was found to reduce emissions, suggesting that transitioning to a service-based economy could serve as a viable strategy for mitigating emissions. This aligns with Hashmi, Hongzhong, Fareed, and Bannya [38], who suggested that the service sector remains a low-polluting sector.

Interestingly, the role of urbanisation in emissions was also highlighted, with findings showing a strong positive association between urban expansion and GHG emissions. This supports prior studies, such as that by Vo, Vo, and Ho [71], and by Huang, Zhang, Deng, Zhao, and Wu [67], which linked rapid urbanisation to increased energy consumption, waste management, construction, and transportation—all of which contribute to higher emissions. Financial development yielded mixed results, showing a positive relationship with CO₂ emissions in certain sectors while contributing to N₂O emission reductions. This aligns with the complicated role of financial development identified by Tran [55] and Nguyen, Duong, and Nguyen [58], who found that, despite financial expansion’s positive influence on growth, it can either drive environmental degradation (through promoting energy consumption) or support sustainability (through enabling investments in clean technology). These findings stress the need to redirect financial systems towards sustainable financing mechanisms, such as green bonds and climate-conscious investments, to ensure economic expansion does not come at the cost of environmental harm. Additionally, trade openness had positive and negative effects on emissions, confirming the dual nature of trade’s environmental impact. While trade enhances resource allocation efficiency and promotes the diffusion of clean technologies [59,60], it can also intensify emissions, particularly in developing countries reliant on pollution-heavy industries [61,63]. The findings suggest that the environmental impact of trade openness is highly context-dependent, reinforcing the need for environmental policies that regulate emissions from export-driven industries.

A critical aspect of this study was the differentiation between developed and developing economies, examined through the interaction variable (SEC*DMY). The results reveal that developed countries exhibit lower emission intensities in industrial and manufacturing sectors, possibly due to stricter environmental regulations, cleaner production technologies, and energy efficiency improvements. However, agriculture in developed countries was found to significantly contribute to CO₂ and N₂O emissions, as aligned with the findings of Balogh and Jámbor [22] and Ghimire, Lin, and Zhuang [10], who highlighted the adverse effects of large-scale commercial farming, including excessive fertiliser use and soil degradation. In contrast, developed countries exhibit sectoral decoupling of emissions, where industrial and manufacturing activities contribute significantly to CO₂ and N₂O emissions, but these are reduced in the service sector. Hence, this study provides valuable insights into how economic activities shape environmental outcomes, emphasising the need for sector- and region-specific, and policy-driven approaches to emissions management.

5. Conclusions

Achieving zero carbon emissions is a crucial objective for mitigating climate change. This goal aligns with international accords, such as the Paris Agreement, and the UN’s SDGs. The transition to a low-carbon economy requires a comprehensive understanding of the factors driving GHG emissions. This study has analysed the economic and structural determinants of CO₂ and N₂O emissions to contribute to informed policy development. The key factors analysed include economic growth, industrialisation, agricultural activity, trade openness, financial development, and urbanisation. The findings indicate that economic growth strongly correlates with CO₂ and N₂O emissions, highlighting the environmental impact of expanding economic activities. The industrial sector remains the primary contributor to CO₂ emissions, particularly in developing countries where manufacturing and energy-intensive industries dominate. In contrast, N₂O emissions in developed countries’ agriculture show a rising trend as its activities are expanded, reflecting the adverse effects of large-scale commercial farming. The findings show that developed countries exhibit stronger environmental performance in the industrial and service sectors, largely due to “leapfrogging” into advanced clean technologies and strict emissions management policies from the lessons of pioneering countries. In contrast, agricultural emissions are disproportionately higher in developed countries, reflecting the mechanised and large-scale nature of modern agricultural operations. Urbanisation and trade openness are also linked to higher CO₂ and N₂O emissions, indicating that population growth in urban centres and increased trade activities contribute to environmental challenges. Natural resource rents positively correlate with both emissions, highlighting that resource-dependent economies tend to experience higher pollution levels. However, the findings further reveal that financial development has a mixed impact on emissions in different sectors, showing its mitigation potential through targeted funding for sectorial projects.

The findings have significant implications for policymakers striving to achieve zero carbon emissions while balancing economic growth. For developed economies, the focus should be on controlling emissions in the agricultural sector by more stringent controls on fertiliser and pesticide use, while encouraging a shift to circular and sustainable agriculture. This includes replacing chemical fertilisers with their organic equivalents, promoting the adoption of precision farming techniques to limit resource waste and reduce NOx emissions, and investing in sustainable land use management systems to limit deforestation and land degradation. In contrast, policies in developing countries should focus more heavily on the industrial and manufacturing sectors, such as through introducing emission control policies in the industrial and manufacturing sectors, raising technological standards, and increasing investment in research and development of clean technologies, including carbon capture and storage, electrification of production processes, and energy recycling. Regardless of their level of development, countries are advised to focus on developing the service sector as a viable strategy to reduce emissions. Similarly, low-carbon transport solutions, such as electric buses and public bicycles, should be integrated, as should smart infrastructure and green building standards to mitigate the environmental impact of rapid urbanisation. In the trade sector, policies should integrate environmental standards into global supply chains, requiring transparency on emissions and material origins, thereby reducing the ecological footprint of international trade. Finally, finance should play a supporting role in the transition to a green economy. Financial instruments are recommended to focus on the industrial, manufacturing, and service sectors where they have the greatest impact. This includes issuing green bonds to raise capital for environmentally friendly projects, establishing transformation funds to support businesses in the investment phase of new technologies, and encouraging the banking and private investment sectors to actively participate in financing green initiatives.

This study has several limitations that future research should address. First, while this study primarily focused on CO₂ and N₂O, this does not negate the environmental relevance of other pollutants, such as CH₄, sulphur dioxide (SO₂), or PM. These pollutants also contribute significantly to climate change and air quality degradation. However, CH₄, SO₂, and PM were predominantly excluded due to data limitations and the maintained analytical focus on the most policy-relevant, globally impactful, persistent, and cumulative GHGs [99]. Second, the reliance on national-level data restricted insights into regional variations, highlighting the need for city-level and subnational studies. Third, the dual role of financial development in limiting and facilitating emissions through investments and support requires empirical validation, particularly regarding the effectiveness of green financing mechanisms. Moreover, green innovation and economic digital transformation should be considered endogenous factors behind sustainable development. Fourth, methodologically, this study relied on secondary data and the GMM technique, which may not have fully captured causality and long-term effects among the observed variables. This method also has limitations in addressing the relationship with spillover effects across units (e.g., pollution diffusion or economic growth, where what happens in one region may affect other regions). Future research should focus on these limitations to provide a more comprehensive study.