From Budgets to Biodiversity: How Fiscal Decentralization Shapes Environmental Sustainability in Pakistan

Abstract

This research contributes to the continuing discussion on the causes of environmental degradation by investigating the impact of fiscal decentralization on environmental sustainability utilizing four measures of environmental sustainability and three measures of fiscal decentralization. The annual data from WDI, OECD, and Global Footprint Network from 1990 to 2023 is analyzed, and the auto regression distributive lag (ARDL) model is employed to calculate long-run estimates. The findings show that fiscal decentralization, technological innovation, population, and other control variables, such as foreign direct investment and trade openness, play important roles in determining environmental sustainability. Composite fiscal decentralization, expenditure, and revenue decentralization lead to decreased environmental sustainability while technological innovation improves environmental sustainability. Furthermore, population, foreign direct investment, and trade openness also negatively affect environmental sustainability. The findings suggest that more resources should be allocated for research and development to save the environment.

1. Introduction

Since the 1997 World Earth Summit in Kyoto, a sustainable atmosphere has become a top priority worldwide. The summit underlined the significance of keeping our planet protected from global environmental disasters. According to the World Bank, 90% of people on the planet are now living with the effects of pollution, land degradation, or water scarcity (World Bank urges fresh push on economic threat of pollution|Reuters). The federal government invests significantly in developmental projects to increase the living standards and well-being of people [1,2], which may cause deterioration of the ecosystem, high pollution, loss of biodiversity, and environmental degradation [3]. Moreover, in developing economies, the race to reduce taxes and offer greater incentives often compromises environmental standards, as states pursue development projects at the expense of environmental quality. Similarly, competing interests of different countries can hinder the implementation of cohesive and stringent national environmental regulations, which may result in the absence of a unified approach. Moreover, in the race for development, developing nations neglect environmental regulation enforcement, adequate monitoring, and accountability to save the long-term interests of their citizens.

Therefore, it is proposed that revenue and expenditure responsibilities and authorities should be given to the lower administrative unit to enhance service quality, as well as environmental quality [4]. The distribution of resources in the lower stratum identifies administrative efficiency, revenue generation, and inclusive growth as key factors for local government success and equity [5]. The theoretical literature argues that fiscal decentralization (FD) can reduce carbon emissions by encouraging local authorities to create more eco-friendly environments, leading to less pollution in devastated areas [6]. However, the empirical literature disagrees on the positive relationship between fiscal decentralization and CO₂ emissions. According to [7] and Yang, Yang [8], fiscal decentralization hurts environmental sustainability. According to Millimet [9] and Cutter and DeShazo [10], fiscal decentralization increases environmental quality in rich countries, while in poor countries, fiscal decentralization deteriorates environmental quality and living standards [11]. In Pakistan, fiscal decentralization remained a key aspect of governance reforms. Fiscal decentralization aims to transfer the rights and authorities to provinces and local bodies to create effective resource management at the regional level [12]. This transformation of revenue or devolution of financial authority to provinces and local bodies may enhance the openness of governance structures. Therefore, in Pakistan, the 18th Amendment to the Constitution in 2010 ensured a significant shift in fiscal powers from the federal government to the provinces, which enabled the local government to make more autonomous decisions on budgetary allocations and expenditures. This decentralization allows local governments to tailor development plans and initiatives based on the specific needs and priorities of their communities and environmental quality. By having greater control over financial resources, provinces and local bodies can invest in infrastructure projects, social services, and economic development programs that align with the unique environmental challenges of their regions [13]. Although fiscal decentralization increases a sense of local ownership and accountability, it also ascribes the decision-makers more direct responsibility for the outcomes of their policies. However, in developing nations, effective implementation of fiscal decentralization always remains a great challenge due to capacity constraints at the local level, uneven administrative capabilities, and issues related to revenue generation [14]. So, in Pakistan, the balance between decentralization and maintaining national unity remains a continuous process, requiring ongoing efforts to strengthen local governance structures, build institutional capacity, and ensure equitable resource distribution for sustainable socio-economic development across the country. Empirical research showed greater discrepancy related to the relationship between fiscal decentralization and environment quality. For instance, the studies that emphasized the positive impact of fiscal decentralization on environment sustainability argued that fiscal decentralization empowered local government with authority and resources, so local government is in a better position to tailor environmental policies according to the region-specific environmental challenges [15]. Moreover, fiscal autonomy grants the local government the incentives to invest in projects that make the environment of the region more pleasant and attractive for investors, tourists, and residents. Similarly, the environment-friendly behavior of local government motivates individual residents to become more active participants in shaping and implementing initiatives that positively affect their environs [16]. Thus, fiscal decentralization not only is beneficial for efficient resource allocation but also enhances local government and individuals’ capacity to respond to emerging environmental challenges.

On the contrary, it has been argued that if fiscal decentralization is not carefully managed, it may exert a negative impact on environmental quality [17,18], specifically, when the local government prioritizes economic development and invites foreign investment without environmental regulation. In the absence of sufficient regulations or oversight, foreign investors may be negligent of environmental standards and adopt outdated technologies and practices that degrade air, water, and soil quality [19]. Moreover, if fiscal decentralization causes unequal distribution of resources among regions, it also aggravates environmental disparities. Similarly, regional policies also increase the risk of fragmented and inconsistent environmental regulations across regions, which create hindrances in coordinated and well-managed efforts to address national environmental issues. Therefore, fiscal decentralization offers sustainable environmental development in the presence of careful governance and regulatory frameworks. Thus, this study aims to investigate the impact of fiscal decentralization on environmental sustainability in the context of Pakistan.

Furthermore, technological innovation plays a significant role in environmental sustainability through innovative and environmentally friendly solutions for socio-economic development [20,21]. Environmentally friendly technologies enhance resource efficiency, reduce environmental impact, and form an environmentally resilient society. Innovations in renewable energy expanded clean energy alternatives diminished reliance on fossil fuels and limited greenhouse gas emissions. Similarly, technological innovation in sustainable agriculture, including precision farming and genetic engineering, enhances food security and contributes to environmental quality through the efficient utilization of resources. Furthermore, environmentally friendly technologies not only facilitate the recycling and repurposing of materials but also help to minimize waste and develop a more sustainable consumption model.

Other important determinants of environmental sustainability found in the literature are population growth, trade, and foreign direct investment [22,23]. The literature documents that the relationship between population growth and environmental quality is a complex and multifaceted issue [24]. On one side, rapid population growth can lead to deforestation due to high demand for housing in societies and increased pollution levels due to high demand for energy, water, and food resources [25]. Therefore, high population growth increases soil degradation, loss of biodiversity, and the emission of greenhouse gases, which ultimately exacerbate climate change. On the other hand, it also has been observed that the growing population fosters innovation and technological advancements through increased demand for sustainable solutions and demand for cleaner technologies, renewable energy sources, and more efficient resource management practices [26]. Similarly, the impact of trade on environmental quality is also complex and multifaceted, with both positive and negative dimensions; for instance, in developing nations, international trade contributes to environmental degradation through the increased movement of goods, leading to higher energy consumption, transportation-related emissions, and habitat destruction associated with production [27], but on the other hand, trade can play a positive role in promoting environmental sustainability through the dissemination of environmentally friendly technologies, fostering innovation and the adoption of cleaner production methods [28]. Keeping in view the importance of technological innovation, population growth, trade, and foreign direct investment, this study modeled them into the main objective, the relationship between fiscal decentralization and environment sustainability. In Pakistan, more than 94% of total revenues are generated at the federal level, and the practice of distributing resources solely based on population has resulted in fiscal imbalances, unequal regional development, and increasing dissatisfaction among provinces. When effectively executed, fiscal decentralization can lead to better resource allocation by tailoring public expenditure to the unique priorities and needs of local communities. Nonetheless, in the absence of strong accountability frameworks, it has the potential to deepen regional inequalities and diminish the quality of public services. Therefore, ensuring transparency and building institutional strength are crucial for making decentralization successful.

Fiscal decentralization is the process by which the central government delegates certain taxation and expenditure responsibilities to local governments and allows them to determine the scale and structure of their budget expenditures by themselves. This enables local governments to participate more actively in social management [29]. Samuelson [30] introduced the concept of fiscal decentralization by ensuring consistent service delivery to all tiers of government, leading to increased efficiency. Fiscal decentralization is a key policy instrument for assessing the effectiveness of local government performance, including social and economic development efforts. Tiebout [31] created a classical theory on fiscal decentralization, stating that local governments deliver superior services to the public, resulting in stable economic efficiency. Later, Bodman, Campbell [32] examined the impact of fiscal decentralization on economic output in Australia and concluded that it is the most effective option for increasing efficiency.

The effects of fiscal decentralization on environmental quality have been comprehensively examined in a few theoretical and empirical studies [33,34]. However, their conclusions are inconsistent. According to Glazer [33] and Levinson [34], fiscal decentralization leads to improved environmental quality. The local government may enact stronger environmental regulations based on the welfare of people related to the improvement of environmental quality criteria to support regional economic development. According to Tiebout [31] and Stigler [35], fiscal decentralization can effectively motivate local governments to provide high-quality environmental services and other public service expenditures to draw residents and resources into their jurisdiction. Consequently, fiscal decentralization helps to improve the quality of the environment.

Local governments can easily gather data and understand the requirements and needs of citizens. Local environmental policies are more favorable to improving environmental quality [36,37]. According to Ji, Umar [38], fiscal decentralization leads to a decrease in CO₂ emissions and thus improves environmental quality. A few studies also found a non-linear relationship between fiscal decentralization and environmental quality [39,40]. This suggests that initially, fiscal decentralization worsens the environment, but later on, it improves it. However, it does not improve environmental quality always, since it reduces economic growth and thus leads to reducing the supply of public goods, i.e., environmental quality [41]. This is coupled with the free-riding tendencies of the government and leads to adversely affecting pollution control and environmental quality [42]. Finally, a study suggests that fiscal decentralization leads to higher pollution in local areas and thus negatively affects environmental quality [43].

Using a balanced panel dataset of seven OECD nations, Khan, Ali [11] examined how fiscal decentralization affects CO₂ emissions and demonstrated how it improves environmental quality. Similarly, Chen, Xu [44] examined the causal relationships between environmental decentralization and local environmental governance. They offered compelling proof that the creation of the Supervision Centers for Environmental Protection greatly encouraged businesses to cut back on pollution from emissions.

On the contrary, in highly decentralized nations, Lingyan, Zhao [17] investigated the asymmetric relationship between carbon emissions, environmental innovation, and fiscal decentralization. They demonstrated that fiscal decentralization considerably reduces carbon emissions only at lower to medium emissions quantiles. According to other research, fiscal decentralization lowers environmental quality. Local governments compete to expand the local economy, draw in investment, create more jobs, collect income taxes, and loosen environmental regulations under the fiscal decentralization system. Therefore, fiscal decentralization fails to improve environmental quality efficiently. Pollution cannot be regulated effectively under low-standard environmental restrictions, which deteriorate the quality of the environment [45,46]. Xia, You [29] analyzed the spatial effects of fiscal decentralization on national carbon emissions, using the Spatial Durbin Model. They concluded that fiscal decentralization increases carbon emissions both inside the region and outside of it. Moreover, carbon emissions will be reduced with the help of environmental decentralization. However, environmental decentralization in nearby areas will raise regional carbon emissions.

Recent studies use macroeconomic techniques, particularly the DSGE (Dynamic Stochastic General Equilibrium) model, to assess environmental challenges. Using a neoclassical model, Hao, Chen [47] examined how fiscal decentralization affects environmental quality. They demonstrated that there is an inverted U-shaped link between fiscal decentralization and GDP per capita. Chan [48] used an E-DSGE model to examine how well carbon taxes and other conventional macroeconomic tools such as fiscal and monetary policies offer reductions in air pollution. They discovered that whereas these measures have the potential to stabilize carbon emissions, there are differences in their underlying mechanisms. Similarly, Zhou and Zhang [49] investigated the effects of fiscal decentralization on high-quality economic development in China. The study employed data from thirty Chinese provinces between 2010 and 2019. They found that fiscal decentralization fosters high-quality economic development, but this effect is mitigated by environmental control. Research indicates that there is a non-linear inverted U-shape relationship between fiscal decentralization and environmental sustainability. The Phan, Jain [50] outlined the impact of fiscal decentralization on emissions in Asian countries and found that increasing revenue and expenditure decentralization reduce CO₂ emissions, while decreasing expenditure decentralization increases CO₂ emissions over time. The empirical data show that the effect of LFD on CO₂ is not symmetric. Finally, Elheddad, Djellouli [51] also found a non-linear association between fiscal decentralization and energy usage in a sample of Chinese regions.

The lack of a universal agreement regarding the effects of fiscal decentralization on the environment is mostly the result of contrasting theoretical and empirical models with poor explanatory capacity. However, the majority of research lacks a thorough talk about how government expenditure is structured under fiscal decentralization; few studies analyze distortions; most just look at average rates.

Previous research on the relationship between technological innovation and ecological footprint has yielded mixed results. Khattak, Ahmad [52] studied the correlation between technological innovation and CO₂ emissions in BRICS nations from 1980 to 2016. The study found innovation increased CO₂ emissions, except in Brazil. Similarly, Usman and Hammar [53] also found that technological innovation harmed environmental sustainability across APEC countries. However, some experts believe that technological innovation has a crucial role in environmental sustainability. Technological innovation can boost productivity and economic development and contribute to environmental sustainability [54]. Technological innovation can enhance energy efficiency [55] and move the energy structure toward clean energy [56], thus stimulating the development of efficient energy markets [57], finally reducing the ecological imprint.

Research Gap

This section is developed to identify the gap in the existing literature and attempts to explain how this study will bridge this gap. First, measuring environmental sustainability is a complex process that involves evaluating the influence of human activities on the environment across several domains, including environmental, social, and economic factors. So, it is essential to consider various dimensions and indicators to measure environmental sustainability [58]. As far as the existing literature is concerned, it utilized traditional measures like carbon emissions (CO₂) [29,47,59,60], sulfur dioxide (SO₂) [47,61], and ecological footprint to measure environmental sustainability [62]. For instance, Li, Younas [63] examined the effect of fiscal decentralization on CO₂ emission, and Ahmed Memon, Ali [62] examined the effect of fiscal decentralization on the ecological footprint in Pakistan. Still, none of these studies used all these proxies simultaneously [29,47,56,59,60,64,65]. Therefore, this study attempted to contribute to the literature and utilized three dimensions individually, namely, CO₂, natural resource depletion, and ecological footprint [47,62,64,65], and this study also developed an environment sustainability index and used it in empirical estimation. Natural resource depletion is an essential aspect, but it is not the only measure of ecological sustainability [65]. The ecological footprint can be viewed as a comprehensive proxy for environmental sustainability [64]. The detail of elements for each dimension is provided in

Table 1. Variable measurement and descriptive analysis.

Variable	Indicator	Measurement	Source	References	Mean	ST.D	Min	Max	JB
EQ	Environment quality	Index	Authors’ calculation	Authors’ calculation	2.940	1.000	1.872	2.132	1.217 (0.544)
LCO2	Carbon emission	CO2 emissions (metric tons per capita)	WDI	[39,40]	−0.355	0.156	−0.681	−0.085	0.992 (0.608)
LEF	Ecological footprints	Global per hectare	Global Footprint Network	[62]	−0.260	0.045	−0.349	−0.167	0.871 (0.646)
LNRD	Natural resource depletion	Natural resources depletion (% of GNI)	WDI		0.213	0.319	−0.464	0.767	1.553 (0.460)
LFD	Fiscal decentralization	FD = decentralized revenue/1-non-decentralized expenditure	Authors’ calculation	[62]	8.441	0.348	6.907	9.465	200 (0.000)
LTI	Technological Innovation	Patent on environmental related technologies	OECD	[62]	2.146	0.998	−0.443	4.615	0.527 (0.768)
LPOP	Population	Population, total	WDI		18.97	0.220	18.56	19.30	2.395 (0.301)
LTRADE	Trade openness	Trade (% of GDP)	WDI	[62]	3.409	0.160	3.066	3.650	1.887 (0.389)
LFDI	Foreign direct investment	Foreign direct investment, net inflows (% of GDP)	WDI	[29]	−0.237	0.553	−1.172	1.110	3.354 (0.186)
LRD	Revenue decentralization		Authors’ calculation		3.407	2.957	0.347	8.647	4.437 (0.108)
LED	Expenditure decentralization		Authors’ calculation		8.151	9.031	0.195	24.814	6.004 (0.049)

. Second, most of the studies used revenue and expenditure decentralization [50], while this study uses three dimensions of fiscal decentralization, namely, revenue decentralization, expenditure decentralization, and composite decentralization (decentralized revenue/1-non-decentralized expenditure). Third, this study contributes to the literature by investigating the impact of population and technological innovation on environmental sustainability in the context of Pakistan.

This paper is structured in multiple sections: after the Introduction Section including the literature review and the Research Gap Section, it includes the Data and Methodology, Results and Discussion, and Conclusions Sections.

2. Data and Methodology

This study analyzes the effects of fiscal decentralization, population growth, and technological innovation along with other controlled variables, trade and foreign direct investment, on environmental quality in Pakistan over the time from 1990 to 2023. Secondary data has been collected from the WDI, OECD, and Global Footprint Network. The period is chosen based on data availability. This study used CO₂ emission (LCO₂), ecological footprint (LEF), natural resource depletion (LNRD), and the environment quality index (LEQ) as dependent variables and the composite fiscal decentralization index (LFD), revenue decentralization (LRD), and expenditure decentralization (LED) as independent variables. This study develops three models, and in each model, four equations are estimated. The general econometric form of the estimated equations in four developed models is provided below:

Model 1

$${\mathrm{LCO}}_2={\beta }_0+{\beta }_{1}L\mathrm{FD}+{\beta }_2\mathrm{LTI}+{\beta }_3\mathrm{LPOP}+{\beta }_4\mathrm{LTRADE}+{\beta }_5\mathrm{L}\mathrm{F}\mathrm{D}\mathrm{I}+\mathsf{\epsilon }$$

(1)

$$\mathrm{LEF}={\beta }_0+{\beta }_{1}L\mathrm{F}\mathrm{D}+{\beta }_2\mathrm{L}\mathrm{T}\mathrm{I}+{\beta }_3\mathrm{L}\mathrm{P}\mathrm{O}\mathrm{P}+{\beta }_4\mathrm{L}\mathrm{T}\mathrm{R}\mathrm{A}\mathrm{D}\mathrm{E}+{\beta }_5\mathrm{L}\mathrm{F}\mathrm{D}\mathrm{I}+\mathsf{\epsilon }$$

(2)

$$\mathrm{LNRD}={\beta }_0+{\beta }_{1}L\mathrm{F}\mathrm{D}+{\beta }_2\mathrm{L}\mathrm{T}\mathrm{I}+{\beta }_3\mathrm{L}\mathrm{P}\mathrm{O}\mathrm{P}+{\beta }_4\mathrm{L}\mathrm{T}\mathrm{R}\mathrm{A}\mathrm{D}\mathrm{E}+{\beta }_5\mathrm{L}\mathrm{F}\mathrm{D}\mathrm{I}+\mathsf{\epsilon }$$

(3)

$$\mathrm{LEQ}={\beta }_0+{\beta }_{1}L\mathrm{F}\mathrm{D}+{\beta }_2\mathrm{L}\mathrm{T}\mathrm{I}+{\beta }_3\mathrm{L}\mathrm{P}\mathrm{O}\mathrm{P}+{\beta }_4\mathrm{L}\mathrm{T}\mathrm{R}\mathrm{A}\mathrm{D}\mathrm{E}+{\beta }_5\mathrm{L}\mathrm{F}\mathrm{D}\mathrm{I}+\mathsf{\epsilon }$$

(4)

Model 2

$${\mathrm{LCO}}_2={\beta }_0+{\beta }_{1}LR\mathrm{D}+{\beta }_2\mathrm{LTI}+{\beta }_3\mathrm{LPOP}+{\beta }_4\mathrm{LTRADE}+{\beta }_5\mathrm{L}\mathrm{F}\mathrm{D}\mathrm{I}+\mathsf{\epsilon }$$

(5)

$$\mathrm{LEF}={\beta }_0+{\beta }_{1}LR\mathrm{D}+{\beta }_2\mathrm{L}\mathrm{T}\mathrm{I}+{\beta }_3\mathrm{L}\mathrm{P}\mathrm{O}\mathrm{P}+{\beta }_4\mathrm{L}\mathrm{T}\mathrm{R}\mathrm{A}\mathrm{D}\mathrm{E}+{\beta }_5\mathrm{L}\mathrm{F}\mathrm{D}\mathrm{I}+\mathsf{\epsilon }$$

(6)

$$\mathrm{LNRD}={\beta }_0+{\beta }_{1}LR\mathrm{D}+{\beta }_2\mathrm{L}\mathrm{T}\mathrm{I}+{\beta }_3\mathrm{L}\mathrm{P}\mathrm{O}\mathrm{P}+{\beta }_4\mathrm{L}\mathrm{T}\mathrm{R}\mathrm{A}\mathrm{D}\mathrm{E}+{\beta }_5\mathrm{L}\mathrm{F}\mathrm{D}\mathrm{I}+\mathsf{\epsilon }$$

(7)

$$\mathrm{LEQ}={\beta }_0+{\beta }_{1}LR\mathrm{D}+{\beta }_2\mathrm{L}\mathrm{T}\mathrm{I}+{\beta }_3\mathrm{L}\mathrm{P}\mathrm{O}\mathrm{P}+{\beta }_4\mathrm{L}\mathrm{T}\mathrm{R}\mathrm{A}\mathrm{D}\mathrm{E}+{\beta }_5\mathrm{L}\mathrm{F}\mathrm{D}\mathrm{I}+\mathsf{\epsilon }$$

(8)

Model 3

$${\mathrm{LCO}}_2={\beta }_0+{\beta }_{1}L\mathrm{ED}+{\beta }_2\mathrm{LTI}+{\beta }_3\mathrm{LPOP}+{\beta }_4\mathrm{LTRADE}+{\beta }_5\mathrm{L}\mathrm{F}\mathrm{D}\mathrm{I}+\mathsf{\epsilon }$$

(9)

$$\mathrm{LEF}={\beta }_0+{\beta }_{1}L\mathrm{E}\mathrm{D}+{\beta }_2\mathrm{L}\mathrm{T}\mathrm{I}+{\beta }_3\mathrm{L}\mathrm{P}\mathrm{O}\mathrm{P}+{\beta }_4\mathrm{L}\mathrm{T}\mathrm{R}\mathrm{A}\mathrm{D}\mathrm{E}+{\beta }_5\mathrm{L}\mathrm{F}\mathrm{D}\mathrm{I}+\mathsf{\epsilon }$$

(10)

$$\mathrm{LNRD}={\beta }_0+{\beta }_{1}L\mathrm{E}\mathrm{D}+{\beta }_2\mathrm{L}\mathrm{T}\mathrm{I}+{\beta }_3\mathrm{L}\mathrm{P}\mathrm{O}\mathrm{P}+{\beta }_4\mathrm{L}\mathrm{T}\mathrm{R}\mathrm{A}\mathrm{D}\mathrm{E}+{\beta }_5\mathrm{L}\mathrm{F}\mathrm{D}\mathrm{I}+\mathsf{\epsilon }$$

(11)

$$\mathrm{LEQ}={\beta }_0+{\beta }_{1}L\mathrm{E}\mathrm{D}+{\beta }_2\mathrm{L}\mathrm{T}\mathrm{I}+{\beta }_3\mathrm{L}\mathrm{P}\mathrm{O}\mathrm{P}+{\beta }_4\mathrm{L}\mathrm{T}\mathrm{R}\mathrm{A}\mathrm{D}\mathrm{E}+{\beta }_5\mathrm{L}\mathrm{F}\mathrm{D}\mathrm{I}+\mathsf{\epsilon }$$

(12)

The subscript t shows the period t = 1990, 1991… 2023. In the above 12 models, only dependent variables are different, while independent variables are the same. In model 1, composite fiscal decentralization is used as an independent variable; in model 2, revenue decentralization is used as the independent variable; and in model 3, expenditure decentralization is used as the independent variable. However, in Equation (1), the dependent variable is carbon emission (LCO₂) used as a measure of environmental sustainability [39,40]. In Equation (2), the ecological footprint is used as the dependent variable to measure environmental sustainability [62]. In Equation (3), the dependent variable is natural resource depletion to measure environmental sustainability [62]. The ecological footprint can be viewed as a comprehensive proxy for environmental sustainability [64]. Finally, in Equation (4), we have made the use of environmental quality index through the method of PCA (principal component analysis) using all three components (carbon emission, ecological footprint, and natural resource depletion) of environmental sustainability. The results of PCA are given in Appendix A (see

Table A6. PCA for the Institutional Quality Index.

Component	Eigenvalue	Difference	Proportion	Cumulative
PC1	1.172	0.215	0.390	0.390
PC2	0.957	0.087	0.319	0.710
PC3	0.869	-	0.289	1.000
Principal Component (Eigenvectors)
Variable	PC1	PC2	PC3	-
EF	0.499	0.789	0.356	-
CO2	−0.654	0.074	0.752	-
NR	0.567	−0.608	0.553	-

Note: EF, CO₂, and NR are ecological footprint, carbon emissions, and natural resource depletion, respectively.

,

Table A7. Correlation between indicators of financial inclusion Fi1 Fi2 Fi3 Fi4.

Correlation	EF	CO2	NR
EF	1.000
CO2	−0.093	1.0000
NR	0.043	−0.116	1.0000

,

Table A8. Kaiser–Meyer–Olkin measure.

KMO

0.526

and

Table A9. ADLS bounds test including structural break.

Statistic	Model 1	Model 2	Model 3	Model 4
F-Statistic	5.650 ***	3.796 **	35.96 ***	3.695 **

Note: The values in parentheses are p-values. ***, and ** show significance at 1% and 5% levels.

). The descriptive statistics are provided in Table 1, while the results of the correlation analysis are provided in

Table 2. Correlation matrix.

	LCO2	LEF	LNRD	EQ	LFD	LRD	LED	LTI	LNFDI	LPOP	LTRADE
LCO2	1
LEF	−0.105	1
LNRD	−0.105	0.026	1
EQ	−0.092	0.999	0.021	1
LFD	−0.139	−0.227	0.059	−0.230	1
LRD	−0.603	−0.089	−0.400	−0.095	−0.020	1
LED	−0.033	−0.276	−0.261	−0.267	0.022	0.370	1
LTI	−0.537	0.238	0.331	0.232	−0.073	0.091	−0.299	1
LFDI	0.086	0.484	0.156	0.482	−0.401	−0.034	−0.293	0.225	1
LPOP	0.940	−0.319	0.052	−0.308	−0.023	−0.702	−0.046	−0.517	−0.099	1
LTRADE	−0.408	0.029	−0.083	0.025	−0.274	0.613	0.022	0.338	0.362	−0.497	1

Source: Authors’ calculations.

.

2.1. Unit Root Test

This study utilized a Zivot–Andrews structural break unit root test to examine the stationarity among the series. Clemente Montañés and Reyes [66] developed this stationarity test that includes two structural breaks in a series, separating it from the ZA test. The Clemente–Montañés–Reyes test compares two models, one using additive outliers and the other using inventive outliers. The results are reported in

Table 3. Zivot–Andrews structural break unit root test.

Variable	t-Statistic	Critical Value	B.P	t-Statistic	Critical Value	B.P	t-Statistic	Critical Value	B.P	t-Statistic	Critical p-Value 5%	B.P
Zivot–Andrews structural break unit root test at level
	With intercept			With trend			Both intercept and trend			Clemente–Montanes–Reyes
EQ	−4.165	−4.80	2010	−4.103	−4.42	2000	−4.181	−5.08	2010	−3.889	−5.490	2007,2019
LCO2	−4.437	−4.80	2010	−4.056	−4.42	2008	−4.288	−5.08	2010	−3.852	−5.490	2003,2015
LEF	−4.141	−4.80	2005	−4.074	−4.42	2000	−4.202	−5.08	2015	−3.868	−5.490	2007,2019
LNRD	−4.504	−4.80	3015	−3.251	−4.42	2005	−2.832	−5.08	2010	−4.006	−5.490	2000,2014
LFD	−5.336	−4.80	2003	−4.779	−4.42	2003	−5.345	−5.08	2005	−0.341	−5.490	2001,2006
LTI	−4.725	−4.80	2003	−2.829	−4.42	2014	−4.350	−5.08	2003	−2.771	−5.490	2003,2014
LPOP	−4.577	−4.58	2012	−3.085	−4.42	2008	−3.034	−5.08	2007	−2.761	−5.490	2001,2011
LTRADE	−3.174	−4.80	2005	−2.968	−4.42	2001	−3.982	−5.08	2005	−2.267	−5.490	1997,2006
LFDI	−3.544	−4.80	2004	−2.629	−4.42	2007	−3.593	−5.08	2004	−4.449	−5.490	2003,2010
LRD	−7.615	−4.80	1998	−3.769	−4.42	2003	−7.485	−5.08	1998	−6.066	−5.490	1999,2017
LED	−3.409	−4.80	2018	−3.792	−4.42	2016	−3.700	−5.08	2015	−4.052	−5.490	2002,2016
Zivot–Andrews structural break unit root test at first difference
EQ	−5.540	−4.80	2002	−5.347	−4.42	2003	−5.760	−5.08	2018	−6.172	−5.490	2005,2015
LCO2	−5.504	−4.80	2014	−5.365	−4.42	1998	−5.478	−5.08	2018	−5.802	−5.490	2008,2015
LEF	−5.537	−4.80	2013	−5.342	−4.42	2012	−5.800	−5.08	2018	−6.094	−5.490	2005,2015
LNRD	−4.315	−4.80	2006	−4.479	−4.42	2016	−4.526	−5.08	2015	−5.633	−5.490	2007,2013
LFD	−6.021	−4.80	2009	−5.137	−4.42	2008	−5.932	−5.08	2009	−8.871	−5.490	2006,2009
LTI	−8.115	−4.80	1998	−8.606	−4.42	2004	−9.048	−5.08	2006	−6.028	−5.490	1999,2003
LPOP	4.861	−4.80	2018	−3.035	−4.42	2018	−5.793	−5.08	2018	−5.555	−5.490	2003,2013
LTRADE	−6.389	−4.80	2002	−5.737	−4.42	1999	−6.974	−5.08	2002	−5.961	−5.490	1998,2006
LFDI	−5.858	−4.80	2008	−5.173	−4.42	2012	−5.502	−5.08	2008	−6.325	−5.490	1998,2007
LRD	−7.261	−4.80	1998	−6.508	−4.42	1999	−7.981	−5.08	2001	−8.700	−5.490	1996,2016
LED	−5.591	−4.80	2001	−5.131	−4.42	2012	−6.000	−5.08	2001	−6.477	−5.490	1997,2016

Note: B.P indicates structural break point. All variables are used in log form.

.

2.2. Cointegration Test

The next phase in the research procedure is to check cointegration among the series. This study used the Johansen cointegration approach because it is extensively used by researchers and is appropriate for this investigation. The Johansen cointegration test is beneficial and produces more reliable results in a multivariate framework. Before using the Johansen cointegration process to calculate the number of cointegrating connections between the independent and dependent variables, it is required to determine whether the variables are integrated as I(0) or I(1). If all variables used in this analysis are integrated with mix order one, I(1) and I(0), the Johansen cointegration test can be used to count the number of cointegrating relationships between the independent and dependent variables [67]. The Gregory and Hansen [68] cointegration test is used to ensure the reliability of cointegration effects. The most commonly used tests, unit root and cointegration, are extensively established in the empirical literature.

2.3. Autoregressive Distributive Lag Model

Based on the unit root test results, it is decided to apply the autoregressive distributed lag model (ARDL) as a primary econometric model. In a time series framework, ARDL has the dynamic advantage of calculating long-run and short-run relationships among variables. The ARDL model incorporates lag values of both dependent and independent variables. By incorporating lag values, the autoregressive distributed lag model acknowledges the presence of endogeneity, serial correlation, and other complications in time series data [69]. For values of both dependent and independent variables, by incorporating lag values, the autoregressive distributed lag model acknowledges the presence of endogeneity, serial correlation, and other complications in time series data [69].

The autoregressive distributed lag model has several advantages; it handles the mix order variables and a small sample size, and it captures long-run and short-run dynamics. The ARDL model, when incorporated into the methodology, can offer a thorough framework for examining the connections between different variables and can reveal details about both their dynamics of equilibrium [70]. Its adaptability and dependability make it a powerful instrument for empirical analysis and help improve the breadth and rigor of research.

Model Specification ARDL

To explain the symmetric relationship among the modeled variable, this study utilized the standard method of ARDL.

Model 1

$$\begin{gathered} ∆L{\mathrm{CO}}_2={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{C}\mathrm{O}2}_{t-1}+\beta }_2{LFD}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+\mathsf{\alpha }}_3{LPOP}_{t-i}+ \\ {{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LFD}_{t-1}+\mathsf{\epsilon } \end{gathered}$$

(13)

$$\begin{gathered} ∆LEF={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{E}\mathrm{F}}_{t-1}+\beta }_2{LFD}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+\mathsf{\alpha }}_3{LPOP}_{t-i}+ \\ {{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LFD}_{t-1}+\mathsf{\epsilon } \end{gathered}$$

(14)

$$\begin{gathered} ∆LNR={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{N}\mathrm{R}\mathrm{D}}_{t-1}+\beta }_2{LFD}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+\mathsf{\alpha }}_3{LPOP}_{t-i}+ \\ {{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LFD}_{t-1}+\mathsf{\epsilon } \end{gathered}$$

(15)

$$\begin{gathered} ∆EQ={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{E}\mathrm{Q}}_{t-1}+\beta }_2{LFD}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+\mathsf{\alpha }}_3{LPOP}_{t-i}+ \\ {{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LFD}_{t-1}+\mathsf{\epsilon } \end{gathered}$$

(16)

Model 2

$$\begin{gathered} ∆L{\mathrm{CO}}_2={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{C}\mathrm{O}2}_{t-1}+\beta }_2{LRD}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+\mathsf{\alpha }}_3{LPOP}_{t-i}+ \\ {{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LRD}_{t-1}+\mathsf{\epsilon } \end{gathered}$$

(17)

$$\begin{gathered} ∆LEF={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{E}\mathrm{F}}_{t-1}+\beta }_2{LRD}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+\mathsf{\alpha }}_3{LPOP}_{t-i}+ \\ {{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LRD}_{t-1}+\mathsf{\epsilon } \end{gathered}$$

(18)

$$\begin{gathered} ∆LNR={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{N}\mathrm{R}\mathrm{D}}_{t-1}+\beta }_2{LRD}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+\mathsf{\alpha }}_3{LPOP}_{t-i}+ \\ {{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LRD}_{t-1}+\mathsf{\epsilon } \end{gathered}$$

(19)

$$\begin{gathered} ∆EQ={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{E}\mathrm{Q}}_{t-1}+\beta }_2{LRD}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+\mathsf{\alpha }}_3{LPOP}_{t-i}+ \\ {{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LRD}_{t-1}+\mathsf{\epsilon } \end{gathered}$$

(20)

Model 3

$$\begin{gathered} ∆L{\mathrm{CO}}_2={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{C}\mathrm{O}2}_{t-1}+\beta }_2{LED}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+\mathsf{\alpha }}_3{LPOP}_{t-i}+ \\ {{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LED}_{t-1}+\mathsf{\epsilon } \end{gathered}$$

(21)

$$\begin{gathered} ∆LEF={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{E}\mathrm{F}}_{t-1}+\beta }_2{LED}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+\mathsf{\alpha }}_3{LPOP}_{t-i}+ \\ {{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LED}_{t-1}+\mathsf{\epsilon } \end{gathered}$$

(22)

$$\begin{gathered} ∆LNR={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{N}\mathrm{R}\mathrm{D}}_{t-1}+\beta }_2{LED}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+\mathsf{\alpha }}_3{LPOP}_{t-i}+ \\ {{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LED}_{t-1}+\mathsf{\epsilon } \end{gathered}$$

(23)

$$\begin{gathered} ∆EQ={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{E}\mathrm{Q}}_{t-1}+\beta }_2{LED}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+\mathsf{\alpha }}_3{LPOP}_{t-i}+ \\ {{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LED}_{t-1}+\mathsf{\epsilon } \end{gathered}$$

(24)

Model 4

$$\begin{gathered} ∆L{\mathrm{CO}}_2={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{C}\mathrm{O}2}_{t-1}+\beta }_2{LFD}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+{\beta }_7{SB}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+ \\ \mathsf{\alpha }}_3{LPOP}_{t-i}+{{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LED}_{t-1}+{\alpha }_7{SB}_{t-i}+\mathsf{\epsilon } \end{gathered}$$

(25)

$$\begin{gathered} ∆LEF={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{E}\mathrm{F}}_{t-1}+\beta }_2{LFD}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+{\beta }_7{SB}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+ \\ \mathsf{\alpha }}_3{LPOP}_{t-i}+{{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LED}_{t-1}+{\alpha }_7{SB}_{t-i}+\mathsf{\epsilon } \end{gathered}$$

(26)

$$\begin{gathered} ∆LNR={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{N}\mathrm{R}\mathrm{D}}_{t-1}+\beta }_2{LFD}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+{\beta }_7{SB}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+ \\ \mathsf{\alpha }}_3{LPOP}_{t-i}+{{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LED}_{t-1}+{\alpha }_7{SB}_{t-i}+\mathsf{\epsilon } \end{gathered}$$

(27)

$$\begin{gathered} ∆EQ={\beta }_0+{{\beta }_1{\mathrm{L}\mathrm{E}\mathrm{Q}}_{t-1}+\beta }_2{LFD}_{t-1}+{{\beta }_3{LTI}_{t-1}+\beta }_4{LTRADE}_{t-1}+ \\ {{\beta }_5{LFDI}_{t-1}+\mathsf{\beta }}_6{{{LPOP}_{t-1}+{\beta }_7{SB}_{t-1}+\alpha }_{1}LCO2}_{t-i}+{{\alpha }_2{LTI}_{t-i}+ \\ \mathsf{\alpha }}_3{LPOP}_{t-i}+{{\alpha }_4{TLTRDE}_{t-i}+\alpha }_5{LFDI}_{t-i}+{\alpha }_6{LED}_{t-1}+{\alpha }_7{SB}_{t-i}+\mathsf{\epsilon } \end{gathered}$$

(28)

In Equations (13) to Equation (28), α0, α1, α2, α3, α4, α5, α6, and α7 represent the long-run relationship, while β1 t0 β7 represent short-run dynamics. In the next step, the study follows the error correction:

$$Y_t={\mathsf{\alpha }}_0+{\beta }_1{X}_{t-1}-\mathsf{\delta }\text{-}{\eta }_{t-1}+{\mathsf{\epsilon }}_t$$

(29)

In Equations (29), the term ${\eta }_{t-1}$ is the error correction term, and δ shows the speed of adjustment; for a stable model, the study expects the negative sign with ECM, and the value of η is less than one.

2.4. Stability Test

In the next step, we applied the stability test. CUSUM and CUSUM of squares are important statistical techniques for quality control and process monitoring. CUSUM measures deviations from a target, detecting changes in the mean or process level, whereas CUSUM of squares analyzes squared deviations to monitor fluctuations in process variability. Both systems use control limits on charts to detect substantial changes in industrial and healthcare processes, allowing for the speedy resolution of deviations from expected behavior [71]. The results are provided in Appendix A.

2.5. Diagnostic Test

The Breusch–Godfrey test, commonly known as the serial correlation LM test, is essential for time series analysis and econometrics. It detects serial correlation in regression model residuals, which might affect model dependability. The test detects autocorrelation by computing an LM statistic and assessing the linear relationship between residuals and lagged values [72]. Moreover, we have also tested heteroscedasticity. The Breusch–Pagan–Godfrey heteroskedasticity test (BPG) is used in econometrics to discover heteroskedasticity in regression models, which might influence results. The test applies an additional regression of squared residuals to the model’s independent variables. The resulting Lagrange multiplier (LM) statistic, calculated using a chi-squared distribution under the null hypothesis of no heteroskedasticity, is critical for verifying regression models and improving the dependability of conclusions in economic and statistical analyses [73]. The results are provided in Appendix A (see

Table A4. Breusch–Godfrey serial correlation LM test.

Model 1
Equation (1)		Equation (2)		Equation (3)		Equation (4)
F-statistic	Obs*R-squared	F-statistic	Obs*R-squared	F-statistic	Obs*R-squared	F-statistic	Obs*R-squared
0.430 (0.948)	12.167 (0.838)	0.529 (0.889)	17.133 (0.703)	0.669 (0.778)	12.336 (0.653)	0.207 (0.956)	1.214 (0.943)
Model 2
Equation (1)		Equation (2)		Equation (3)		Equation (4)
F-statistic	Obs*R-squared	F-statistic	Obs*R-squared	F-statistic	Obs*R-squared	F-statistic	Obs*R-squared
0.807 (0.614)	8.921 (0.542)	0.730 (0.676)	7.339 (0.601)	0.448 (0.930)	8.741 (0.847)	0.335 (0.929)	2.834 (0.899)
Model 3
Equation (1)		Equation (2)		Equation (3)		Equation (4)
F-statistic	Obs*R-squared	F-statistic	Obs*R-squared	F-statistic	Obs*R-squared	F-statistic	Obs*R-squared
0.559 (0.815)	5.926 (0.747)	0.714 (0.742)	13.93 (0.603)	0.619 (0.813)	10.89 (0.694)	0.397 (0.922)	4.489 (0.877)

and

Table A5. Heteroskedasticity test: Breusch–Pagan–Godfrey.

Model 1
Equation (1)		Equation (2)		Equation (3)		Equation (4)
F-statistic	Obs*R-squared	F-statistic	Obs*R-squared	F-statistic	Obs*R-squared	F-statistic	Obs*R-squared
0.430 (0.948)	12.167 (0.838)	0.529 (0.889)	17.133 (0.703)	0.669 (0.778)	12.336 (0.653)	0.207 (0.956)	1.214 (0.943)
Model 2
Equation (1)		Equation (2)		Equation (3)		Equation (4)
F-statistic	Obs*R-squared	F-statistic	Obs*R-squared	F-statistic	Obs*R-squared	F-statistic	Obs*R-squared
0.807 (0.614)	8.921 (0.542)	0.730 (0.676)	7.339 (0.601)	0.448 (0.930)	8.741 (0.847)	0.335 (0.929)	2.834 (0.899)
Model 3
Equation (1)		Equation (2)		Equation (3)		Equation (4)
F-statistic	Obs*R-squared	F-statistic	Obs*R-squared	F-statistic	Obs*R-squared	F-statistic	Obs*R-squared
0.559 (0.815)	5.926 (0.747)	0.714 (0.742)	13.93 (0.603)	0.619 (0.813)	10.89 (0.694)	0.397 (0.922)	4.489 (0.877)

).

3. Results and Discussion

This section of the research discusses the outcomes of the research. Table 2 demonstrates the correlation matrix, and it is evident that there is no problem of multicollinearity between the variables of the study. Table 3 shows the results of the Zivot–Andrews and Clemente–Montanes–Reyes structural-break unit root tests. The results show that variables are not stationary at the level, as seen by lower t statistics than critical values. However, all variables are stationary at the first difference except LFD, which is stationary in level form. The calculated values of t statistics for EQ, LCO2, LEF, LNRD, LTI, LPOP, LTRADE, and LFDI are greater than critical values, refuting the null hypothesis of the unit root.

The integrated series at the first difference I(1) allows us to proceed with the use of cointegration, specifically Gregory–Hansen cointegration, to study the long-term relationship between variables. Johansen cointegration (see

Table 4. Results of Johansen cointegration test.

Hypothesis	Trace Statistics	5% Critical Value	Max-Eigen Statistics	5% Critical Value
None *	483.890 ***	197.370	140.725 ***	58.433
At most 1 *	343.165 ***	159.529	91.824 ***	52.362
At most 2 *	251.340 ***	125.615	74.050 ***	46.231
At most 3 *	177.290 ***	95.753	62.381 ***	40.077
At most 4 *	114.909 ***	69.818	48.948 ***	33.876
At most 5 *	65.960 ***	47.856	35.193 ***	27.584
At most 6 *	30.767 **	29.797	19.874 *	21.131
At most 7	10.892	15.494	8.678	14.264

Note: ***, **, and * indicate rejection of null at 1%, 5% and 10% level of significance.

) is primarily used to study the long-term relationship between variables. However, the traditional cointegration test does not consider the presence of a structural break in the long-run relationship, which is a weakness in the reliability of estimated results. To solve this issue, achieve more credible estimates in the presence of such structural breaks, and for a robustness check, we used the GH (Gregory–Hansen) cointegration test. This test considers the endogenous structural breaks in the cointegrating variable.

Table 5. Gregory–Hansen test for cointegration (Model 1–4).

Statistic	Model 1	Model 2	Model 3	Model 4
ADF	−5.89 *	−7.50 **	−7.64 **	−7.47 ***
Zt	−5.99	−7.62 **	−8.22 **	−7.59 ***
B.P	2009	2013	2014	2013

Note: ***, **, and * show significance at 1%, 5%, and 10% level of significance

clearly shows the presence of cointegration and structural fractures in the model. Pakistan experienced a substantial structural transformation in 2013. GH test results for the remaining models are provided in Appendix A (see

Table A10. Gregory–Hansen test for cointegration (Model 5–12).

Statistic	Model 5	Model 6	Model 7	Model 8	Model 9	Model 10	Model 11	Model 12
ADF	−6.54	−7.50 **	−8.17 ***	−6.27	−7.71 ***	−6.72 *	−8.16 ***	−6.92 ***
Zt	−6.64 *	−7.62 **	−8.67 ***	−6.25	−7.83 ***	−7.16 **	−8.49 ***	−7.26 ***
B.P	2015	2013	2014	2000	2008	2007	2010	2002

Note: ***, **, and * show significance at 1, 5, and 10% levels.

).

The findings of the ARDL bound test are shown in

Table 6. ARDL bounds test.

Equations	Null Hypothesis: No Long-Run Relationships Exist
	Model 1; K = 5	Model 2; K = 5	Model 3; K = 5	Model 3; K = 5
	F-Statistic	F-Statistic	F-Statistic	F-Statistic
Equation (1)	6.186	13.641	22.862	6.431
Equation (2)	7.082	21.011	60.521	17.452
Equation (3)	5.082	13.075	40.775	20.271
Equation (4)	8.031	25.192	34.283	16.042
Critical Value Bounds
Significance	I0 Bound		I1 Bound
10%	2.08		3.00
5%	2.39		3.38
1%	3.06		4.15

, which explains the long-run relationship between modeled variables. The F-statistics for four equations in four models are greater than the given lower and upper bound values, so the null hypothesis, no long-run cointegration, is rejected, as well as the existence of cointegration and the existence of a long-run relationship among environmental quality, fiscal decentralization, and other independent variables of the study.

Long-run results are reported in

Table 7. ARDL long-run results.

Model 1
Equations	Dependent Variables	Independent Variables
		LFD	LTI	LFDI	LPOP	LTRADE	CointEq(−1) *
Equation (1)	LCO2	0.157 ** (0.042)	−0.056 *** (0.00)	0.092 *** (0.00)	0.154 ** (0.046)	0.238 ** (0.019)	−0.191
Equation (2)	LEF	0.111 ** (0.035)	−0.021 ** (0.036)	0.134 *** (0.00)	0.192 ** (0.018)	0.112 ** (0.028)	−0.59
Equation (3)	LNRD	0.086 ** (0.03)	−0.264 *** (0.00)	0.356 *** (0.00)	0.114 *** (0.00)	0.464 ** (0.018)	−0.979
Equation (4)	EQ	2.463 ** (0.045)	−0.489 ** (0.041)	2.961 *** (0.00)	4.246 ** (0.023)	2.501 ** (0.034)	−0.599
Model 2
	D.V	LRD	LTI	LFDI	LPOP	LTRADE	CointEq(−1) *
Equation (1)	LCO2	0.002 (0.405)	−0.039 ** (0.01)	0.103 *** (0.00)	0.320 ** (0.016)	0.170 ** (0.039)	−0.892
Equation (2)	LEF	0.003 ** (0.038)	−0.011 ** (0.089)	0.094 *** (0.00)	0.024 * (0.066)	0.219 *** (0.006)	−0.757
Equation (3)	LNRD	0.022 ** (0.041)	−0.280 *** (0.001)	0.256 *** (0.009)	2.702 *** (0.00)	0.949 ** (0.018)	−0.677
Equation (4)	EQ	0.084 ** (0.034)	−0.276 * (0.075)	2.100 *** (0.00)	0.841 *** (0.009)	4.928 *** (0.007)	−0.607
Model 3
	D.V	LED	LTI	LFDI	LPOP	LTRADE	CointEq(−1) *
Equation (1)	LCO2	0.010 ** (0.046)	−0.070 *** (0.00)	0.129 *** (0.00)	0.024 ** (0.023)	−0.049 (0.442)	−0.687
Equation (2)	LEF	0.012 * (0.097)	−0.006 (0.544)	0.056 ** (0.046)	0.017 (0.136)	0.206 ** (0.047)	−0.98
Equation (3)	LNRD	0.016 ** (0.053)	−0.398 *** (0.00)	0.201 *** (0.022)	3.704 *** (0.00)	0.294 * (0.065)	−0.722
Equation (4)	EQ	0.155 ** (0.036)	−0.127 (0.425)	0.897 ** (0.017)	0.300 ** (0.032)	1.950 *** (0.042)	−0.795
Model 4
	D.V	LED	LTI	LFDI	LPOP	LTRADE	Structural break
Equation (1)	LCO2	0.488 *** (0.001)	−0.116 *** (0.000)	0.220 *** (0.000)	−0.217 *** (0.001)	−0.192 *** (0.007)	0.292 *** (0.002)
Equation (2)	LEF	0.023 ** (0.071)	−0.021 * (0.069)	0.199 *** (0.000)	0.136 *** (0.000)	0.052 *** (0.001)	0.042 ** (0.048)
Equation (3)	LNRD	0.016 * (0.081)	0.263 *** (0.000)	14.82 *** (0.000)	0.489 *** (0.000)	−0.124 ** (0.014)	−0.543 *** (0.000)
Equation (4)	EQ	0.017 * (0.097)	−0.448 * (0.083)	4.299 *** (0.000)	3.040 *** (0.000)	1.165 *** (0.001)	0.888 * (0.061)

Note: Values in parentheses are p-values. ***, **, and * show significance at 1, 5, and 10% levels.

. All variables are highly significant as evident from significant probability values. In model 1, fiscal decentralization is the main independent variable and has an exacerbating effect on carbon emissions. A 1% increase in LFD leads to a 0.15%, increase in LCO₂. The coefficient of LTI reveals that a one percent increase leads to a decrease in LCO₂ by 0.05%, while LFDI, LPOP, and LTRADE all have a positive effect on LCO₂ (see Table 7). It suggests that a 1% increase in these factors leads to decreases in LCO₂ by 0.15%, 0.09%, and 0.23%.

In the remaining three models, again, all the variables have a positive effect on LCO₂ except LTI. TI coefficient values are negative, indicating a pollution-mitigation impact. A 1% increase in LTI significantly decreases LEF, LNRD, and EQ by 0.02%, 0.26%, and 0.48%, respectively (see Table 7). A 1% increase in LFD leads to 0.11%, 0.08%, and 2.46% increases in LEF, LNRD, and EQ, respectively. LTRADE has a statistically significant positive impact on LEF, LNRD, and EQ. A 1% increase in LTRADE raises LEF, LNRD, and EQ by 0.112%, 0.464%, and 2.501%, respectively.

The long-run estimates for the four equations in model 2 reveal that revenue decentralization (LRD) has an insignificant role in determining carbon emission. While it positively determines LEF, LNRD, and EQ, a 1 percent rise in revenue decentralization leads to increases in LEF, LNRD, and EQ by 0.003%, 0.02%, and 0.08%, respectively. Similarly, LTI, LFDI, LPOP, and LTRADE have significant positive effects on LEF, LNRD, and EQ, thus deteriorating environmental quality. However, LTI hurts all three indicators of environmental quality and the EQ index; therefore, it is favorable for improving environmental sustainability.

Finally, the outcomes of model 3 demonstrate that expenditure decentralization (LED) proves to be unfavorable for environmental sustainability. If expenditure decentralization (LED) increases by 1 percent, LCO₂ rises by 0.01%, and LEF, LEF, and LNRD rise by 0.01%, while EQ rises by 0.15%. Similarly, LFDI, LPOP, and LTRADE all have significant positive effects on LCO₂, LEF, LNRD, and EQ and deteriorate environmental quality.

Moreover, the values of ECM in model 1 indicate the rate at which the dependent variable adapts to its long-term relationship with the independent variables following a shock or divergence from equilibrium. In model 1, cointEq (ECM) shows that model 1 converges toward long-run equilibrium with the speeds of 19% in Equation (1), 59% in Equation (2), 97% in Equation (3), and 59% in Equation (4). The speed of convergence is high in Equation (3), model 1. Similarly, in model 2, the speed of convergence is 89% in Equation (1), 75% in Equation (2), 67% in Equation (3), and 60% in Equation (4). In model 3, the speed of convergence is 68% in Equation (1), 98% in Equation (2), 72% in Equation (3), and 79% in Equation (4). At last, after testing for structural breaks in the data series, we found that Pakistan has faced significant structural change from 2003 to 2018, especially in the years of 2003, 2010, 2015, and 2018 (see Table 3). Model 4 in Table 7 reveals the long-run results after regressing LCO₂, LEF, LNRD, and EQ on a dummy of structural break. The results advocate that all variables are highly significant as indicated by significant p-values. S.B. (structural break) is also significant in all of the four equations. The 18th Constitutional Amendment, passed in 2010, made significant modifications to Pakistan’s federal system. It gave provinces additional authority over resources like natural gas and minerals. The amendment abolished the president’s unilateral ability to dissolve Parliament, transforming Pakistan from a semi-presidential to a parliamentary republic. It also handed self-governing, legislative, and financial autonomy to provincial authorities, dissolving key federal ministries and empowering provinces with valid constitutional rights in controlling and exploiting their natural resources. This amendment aimed to enhance fiscal decentralization by empowering provincial governments [74]. The detailed results of the short run period for models 1, 2, and 3 are provided in Appendix A (

Table A1. ARDL short run results (model 1).

Variable	(DV = LOC2)	(DV = LEF)	(DV = LNRD)	(DV = EQ)
LFD	0.053 *** (0.001)	−0.083 *** (0.000)	0.075 ** (0.014)	1.864 ** (0.028)
LTI	-	−0.018 ** (0.038)	−0.123 *** (0.000)	−0.400 ** (0.032)
LFDI	0.039 *** (0.002)	0.060 *** (0.000)	0.278 *** (0.000)	1.309 *** (0.002)
LPOP	6.247 * (0.075)	9.226 *** (0.001)	19.760 *** (0.009)	435.991 ** (0.019)
LTRADE	−0.113 ** (0.047)	−0.197 *** (0.000)	0.928 *** (0.000)	4.364 *** (0.004)
CointEq(−1) *	−0.191	−0.590	−0.979	−0.599

Note: The values in parentheses are p-values. ***, **, and * show significance at 1, 5, and 10% levels.

,

Table A2. ARDL Short Run Results (Model 2).

Variable	(DV = LCO2)	(DV = LEF)	(DV = LNRD)	(DV = EQ)
LRD	0.004 (0.024)	0.005 ** (0.038)	0.013 (0.128)	0.123 ** (0.035)
LTI	0.049 ** (0.015)	−0.016 (0.116)	−0.337 *** (0.004)	−0.405 (0.101)
LFDI	0.131 *** (0.003)	0.139 *** (0.000)	0.167 ** (0.028)	3.082 *** (0.000)
LPOP	19.70 *** (0.004)	0.035 *** (0.071)	3.247 *** (0.001)	1.235 ** (0.014)
LTRADE	−0.125 (0.102)	−0.323 ** (0.011)	1.140 ** (0.010)	−6.975 * (0.006)
CointEq(−1) *	−0.892	−0.757	−0.677	−0.607

Note: The values in parentheses are p-values. ***, **, and * show significance at 1, 5, and 10% levels.

and

Table A3. ARDL short run results (model 3).

Variable	(DV = LCO2)	(DV = LEF)	(DV = LNRD)	(DV = EQ)
LED	0.008 ** (0.038)	−0.005 * (0.083)	0.085 *** (0.000)	−0.467 (0.134)
LTI	−0.053 *** (0.005)	−0.037 * (0.095)	−0.051 (0.262)	−0.384 (0.409)
LFDI	0.099 *** (0.000)	0.063 *** (0.000)	0.132 ** (0.050)	2.703 *** (0.002)
LPOP	0.019 ** (0.041)	9.822 *** (0.004)	22.89 *** (0.000)	0.903 (0.134)
LTRADE	−0.038 ** (0.452)	−0.288 *** (0.001)	0.881 ** (0.013)	−2.908 ** (0.023)
CointEq(−1) *	−0.687	−0.980	−0.722	−0.795

Note: The values in parentheses are p-values. ***, **, and * show significance at 1, 5, and 10% levels.

).

Fiscal decentralization increases, and governments may decrease environmental rules to attract higher tax-paying and high-emitting firms to their authorities. This increases the region’s carbon output [29]. Expenditure decentralization leads to increased LCO₂, LEF, LNRD, and EQ and suggests the negative effect on environmental sustainability. Inadequate waste management, inadequate planning, and insufficient oversight can increase environmental degradation [75]. Research indicates a trend toward novel and greener energy sources [76]. Governments and factories are investing much in research and development to find effective solutions [77]. Clean and sustainable energy promotes a green environment and economic development. Our results are in line with the findings of Ahmed Memon, Ali [62] but contradict Awosusi, Adebayo [78], who claim that LTI uses energy sources to improve economic activities, which leads to an increase in LEF rather than decreasing it.

Moreover, population, foreign direct investment, and trade openness all have positive effects on LCO₂, LEF, LNRD, and EQ. The positive coefficient of LFDI indicates that local governments may have relaxed environmental access thresholds to encourage LFDI, leading to higher carbon emissions in the county. Local governments should strictly encourage foreign investment and promote energy-efficient, emission-reducing, and clean production initiatives. These findings are consistent with Xia, You [29]. Finally, LPOP has a positive impact on LCO₂. The growing population leads to increased demand for energy resources. This strains finite resources and may negatively affect environmental quality results. This corroborates the findings of Huo and Peng [59].

The LTRADE outcomes support the outcomes of Wang and Wang [79] and Hashmi, Hongzhong [80], indicating a significant link between LTRADE and environmental quality. Recent research indicates that trade density is the primary driver of increased emissions [81]. Trade openness can influence CO₂ emissions and ecological footprints in several ways, including scale, technological, and composition effects. The scale effect implies that increased trade can increase production and energy consumption, resulting in larger CO₂ emissions. The composition impact refers to the reassignment of resources and traded commodities, which can increase or decrease pollution. The technology effect indicates that trade openness can facilitate the diffusion of cleaner technologies, resulting in lower environmental degradation. When the scale effect outweighs the technology effect, trade leads to increased LCO₂ emissions [82]. Finally, trade openness results in economic globalization, which has the potential to influence natural resource extraction techniques through the efficient transfer of technology [83]. Furthermore, trade openness can influence resource depletion through the same three effects, i.e., scale, composition, and technology. The scale effect implies that increased trade can increase productivity and energy consumption, resulting in greater resource depletion [82].

4. Conclusions

This research explored how fiscal, revenue, and expenditure decentralization influence environmental sustainability in Pakistan over the period 1990–2023. Using indicators such as carbon emissions, ecological footprint, natural resource depletion, and an environmental quality index, the findings reveal a mixed picture; fiscal decentralization is linked to weaker environmental outcomes, while technological innovation has a positive and corrective effect. At the same time, population growth, foreign direct investment, and trade openness tend to place additional stress on the environment. In light of these results, three policy directions stand out. First, decentralization strategies should be carefully designed to embed environmental safeguards, clarify responsibilities across government tiers, and ensure strong enforcement and monitoring. Second, Pakistan needs to scale up investment in renewable energy, energy-efficient infrastructure, and other green innovations that can counterbalance the environmental costs of growth. Third, strengthening population management and resource-efficient practices (through accessible family planning services and public awareness) can help ease ecological pressures. Taken together, a balanced approach that blends cautious fiscal decentralization with a strong commitment to innovation and environmental responsibility offers a practical pathway toward a more sustainable future for Pakistan.