Economic Growth, FDI, Tourism, and Agricultural Productivity as Drivers of Environmental Degradation: Testing the EKC Hypothesis in ASEAN Countries

Abstract

This study examines the long-run relationship between carbon dioxide (CO₂) emissions and key macroeconomic and sectoral drivers in ten ASEAN economies from 1995 to 2023. Employing Driscoll–Kraay standard errors, Prais–Winsten regression, heteroskedastic panel-corrected standard errors, Fully Modified Ordinary Least Squares (FMOLS), Dynamic Ordinary Least Squares (DOLS), and Canonical Cointegrating Regression (CCR) estimators, the analysis accounts for cross-sectional dependence, slope heterogeneity, and endogeneity. Results indicate that GDP exerts a more-than-unitary positive effect on emissions, with a negative GDP-squared term supporting the Environmental Kuznets Curve. Agriculture raises emissions through land-use change and high-emission cultivation practices, while tourism shows a negative association likely reflecting territorial accounting effects. Trade openness increases emissions, highlighting the carbon intensity of export structures, whereas foreign direct investment exerts no significant net effect. These results suggest that ASEAN economies must accelerate renewable energy adoption, promote climate-smart agriculture, embed enforceable environmental provisions in trade policy, and implement rigorous sustainability screening for FDI to achieve low-carbon growth trajectories.

1. Introduction

Over the past three decades, Southeast Asia has experienced rapid structural transformation, propelled by accelerated industrialization, urbanization, and deepening regional economic integration. This process has positioned the Association of Southeast Asian Nations (ASEAN)—comprising Brunei Darussalam, Cambodia, Indonesia, Lao PDR, Malaysia, Myanmar, the Philippines, Singapore, Thailand, and Viet Nam—as one of the most dynamic and fastest-growing regional blocs globally. As of 2025, ASEAN’s combined population exceeds 690 million, and its aggregate gross domestic product (GDP) accounts for approximately 7.3% of global output. The region’s economic dynamism has been sustained by robust growth in trade flows, foreign direct investment (FDI), and sectoral productivity. However, these achievements have been accompanied by significant environmental costs, including rising carbon dioxide (CO₂) emissions, recurrent transboundary haze events, biodiversity loss, land degradation, and accelerated depletion of natural resources [1].

A primary driver of these environmental pressures is ASEAN’s heavy dependence on fossil fuels. Coal and petroleum continue to dominate the energy mix for industrial production, electricity generation, and transportation. Since 2000, Southeast Asia’s energy demand has more than doubled, and without substantial investment in renewable energy, fossil fuel consumption is projected to continue increasing through 2035. This reliance not only exacerbates CO₂ emissions but also exposes the region to global energy price volatility and the economic risks associated with delayed decarbonization. While the global literature on the energy–environment–growth nexus is extensive [2,3], much of the ASEAN-focused research relies on aggregate energy consumption or GDP measures, neglecting sectoral heterogeneity and interlinkages that shape environmental outcomes over time.

Environmental pressures in ASEAN extend well beyond energy-related emissions. Agriculture, forestry, and fishing remain significant contributors to greenhouse gas (GHG) emissions, largely through large-scale plantation expansion, slash-and-burn land clearing, and peatland drainage. In Indonesia and Malaysia, conversion of peatlands to oil palm plantations releases substantial volumes of stored carbon, while biomass burning during land preparation produces severe transboundary haze episodes with detrimental effects on public health, transportation, and biodiversity [4]. These practices also degrade soil quality, reduce water retention capacity, and increase vulnerability to climate extremes. International tourism plays a similarly dual role in ASEAN economies. While generating millions of jobs and substantial foreign exchange earnings, rapid tourism growth has also intensified environmental pressures in popular destinations, particularly in Thailand, Indonesia, the Philippines, and Viet Nam. Coastal and marine ecosystems face degradation from coral reef damage, beach erosion, and increased marine litter [5]. Tourism-related infrastructure and resort operations place additional strain on local resources. In particular, the high demand for energy and water frequently exceeds the capacity of existing waste management systems [6]. The sector’s carbon footprint is further amplified by emissions from air travel, cruise ships, and energy-intensive hospitality facilities. Trade openness, another hallmark of ASEAN’s economic trajectory, has reinforced carbon-intensive production structures. Export portfolios dominated by electronics, textiles, and processed agricultural goods rely heavily on fossil fuel-based energy inputs and transport logistics [7]. These sectoral dynamics are deeply interlinked. Tourism can heighten both energy and water demand; agricultural expansion exacerbates water scarcity while contributing to emissions; and trade-driven industrial activity reinforces fossil fuel dependency. Addressing these interconnected pressures requires an analytical approach that moves beyond aggregated economic indicators to capture sector-specific drivers of environmental degradation.

ASEAN’s socio-economic heterogeneity further complicates this challenge. High-income members such as Singapore and Brunei benefit from advanced infrastructure, service-oriented economies, and comparatively strong environmental governance, resulting in lower emissions intensity per unit of GDP. In contrast, lower-middle-income economies such as Cambodia, Lao PDR, and Myanmar remain dependent on resource-based industries and carbon-intensive manufacturing, with both land and water productivity lagging behind international best practices. This diversity produces divergent environmental outcomes and suggests that ASEAN members are unlikely to follow a uniform development–environment trajectory.

The Environmental Kuznets Curve (EKC) hypothesis, which posits an inverted U-shaped relationship between per capita income and environmental degradation [8], has been widely applied to assess the growth–environment nexus. However, evidence for developing economies, including ASEAN members, remains inconclusive. Many studies indicate that environmental degradation continues to rise beyond the predicted turning point, especially where institutional capacity is weak, technological adoption is slow, and economies remain structurally reliant on carbon-intensive sectors [9,10,11]. This suggests that achieving the EKC turning point in ASEAN will require policies that explicitly account for sectoral patterns, cross-country heterogeneity, and the interaction of multiple environmental drivers.

Despite growing interest in ASEAN’s environmental challenges, notable research gaps persist. First, a significant portion of the literature aggregates economic activity into a single GDP measure, obscuring the distinct environmental impacts of agriculture, tourism, trade, and industrial production. This limits the precision of policy recommendations. Second, relatively few studies employ econometric techniques capable of addressing cross-sectional dependence, slope heterogeneity, and endogeneity—issues that, if ignored, can bias long-run estimates in multi-country settings. Third, there is limited research integrating multiple sectoral drivers into a unified dynamic panel framework that can capture both short-run adjustments and long-run equilibrium relationships between CO₂ emissions, economic growth, and sectoral activities.

This study seeks to address these gaps by examining the dynamic effects of economic growth, agriculture, tourism, trade, and FDI on environmental degradation in ASEAN countries. It makes six key contributions to the literature. First, it adopts a sectoral disaggregated analytical framework, moving beyond the traditional aggregate growth–emissions nexus to capture differentiated environmental impacts of distinct economic sectors. Second, it employs multiple robust econometric techniques—including Driscoll–Kraay standard errors, Prais–Winsten regression with panel-corrected standard errors, Fully Modified Ordinary Least Squares (FMOLS), and Canonical Cointegrating Regression (CCR)—to account for cross-sectional dependence, slope heterogeneity, serial correlation, and potential endogeneity, thereby enhancing the reliability of long-run parameter estimates. Third, by incorporating both GDP and GDP-squared terms, the study explicitly tests for the EKC hypothesis within a multi-sector context, offering fresh evidence on the nonlinear growth–environment relationship in ASEAN. Fourth, the analysis draws on a comprehensive panel dataset spanning nearly three decades (1995–2023), which captures both short-run adjustments and long-run equilibrium relationships, enabling insights into the persistence and dynamics of sectoral impacts on emissions. Fifth, it integrates sector-specific variables, such as agriculture, tourism, and trade, into the EKC framework, thus extending its applicability beyond aggregate indicators and highlighting the heterogeneity of emission drivers in a developing regional bloc. Sixth, by embedding the empirical findings within ASEAN’s climate governance and sustainable development agenda, the study delivers targeted, evidence-based policy recommendations that balance economic competitiveness with environmental stewardship, offering direct relevance for regional policymakers, international development agencies, and climate negotiators.

The findings have direct policy relevance for ASEAN’s environmental governance, currently shaped by frameworks such as the 1997 Regional Haze Action Plan (RHAP) and the 2002 ASEAN Agreement on Transboundary Haze Pollution (AATHP). Despite their importance, these mechanisms have been hindered by delayed ratification—Indonesia only ratified the AATHP in 2014—uneven enforcement, and entrenched sectoral interests [12,13]. By identifying sector-specific environmental drivers, this study offers a foundation for differentiated policy strategies that reflect the structural diversity of ASEAN economies, enabling more targeted and effective interventions for green growth and long-term environmental resilience.

The remainder of this study is organized as follows. Section 2 provides a comprehensive review of the existing literature on the relationship between economic growth and environmental degradation, with a focus on sector-specific drivers such as agriculture, tourism, and foreign direct investment in the ASEAN context. Section 3 describes the data sources, variable definitions, and econometric methodologies employed, including FMOLS and CCR techniques. Section 4 presents empirical results, discusses their implications, and examines the robustness of the findings. Section 5 highlights the policy suggestion developed by empirical findings of this study. Finally, Section 6 concludes the study by summarizing key insights and offering policy recommendations for sustainable development in ASEAN nations.

2. Literature Review

Over the past three decades, the nexus between economic development and environmental sustainability has become one of the most scrutinized topics in environmental economics and policy research. Rapid globalization, industrialization, and sectoral transformation have simultaneously driven economic progress and intensified ecological pressures, particularly in emerging and transitional economies [14]. Among these, ASEAN presents a compelling case for empirical investigation. The region’s economies have experienced sustained GDP growth, deep integration into global value chains (GVCs), rapid expansion in tourism, manufacturing, and agriculture, and substantial inflows of FDI. Yet, these structural transformations have also contributed to a sharp rise in greenhouse gas (GHG) emissions, with CO₂ emissions accounting for the largest share.

ASEAN’s energy profile—dominated by fossil fuels—combined with varying institutional capacities and environmental regulations, creates an uneven landscape in which the growth–emissions relationship can differ substantially between member states. This heterogeneity makes ASEAN a critical testing ground for hypotheses such as the EKC, the pollution haven hypothesis (PHH), and the pollution halo hypothesis, as well as for assessing the role of sectoral composition and trade integration in shaping environmental outcomes. The environmental impact of FDI is often framed in terms of two competing hypotheses [15]. The PHH argues that multinational firms may relocate pollution-intensive production to countries with relatively weaker environmental regulations, thereby increasing host-country emissions. In contrast, the pollution halo hypothesis suggests that FDI can promote cleaner production by transferring advanced technologies, managerial practices, and environmental standards from developed to developing economies, thus contributing to emission reductions. Furthermore, the bloc’s commitment to the Paris Agreement and its own ASEAN Plan of Action for Energy Cooperation (APAEC) underscores the urgency of understanding the determinants of CO₂ emissions from both a theoretical and policy-oriented perspective.

In this context, a growing body of empirical research has examined how macroeconomic drivers, including GDP growth, trade openness, industrial activity, FDI inflows, tourism, and agricultural productivity, interact with CO₂ emissions in ASEAN economies. The literature reflects methodological diversity, ranging from panel cointegration approaches and error correction models to advanced estimators addressing cross-sectional dependence, slope heterogeneity, and endogeneity. Despite this, results remain mixed and context-dependent, reflecting the complex interplay of economic structure, policy frameworks, technological adoption, and energy use patterns across countries and sectors.

The following literature review synthesizes existing empirical findings on the relationships between these macroeconomic and sectoral drivers and CO₂ emissions in ASEAN. By integrating evidence from ASEAN-focused and comparative international studies, it identifies consistent patterns, methodological innovations, and persistent gaps in the literature, particularly regarding sectoral heterogeneity, investment composition, and the role of institutions and clean technologies in mediating environmental impacts.

The relationship between economic growth and CO₂ emissions has been extensively examined within the framework of the EKC, which postulates that environmental degradation initially intensifies with rising income but subsequently declines after surpassing a certain economic threshold [16]. ASEAN economies, characterized by rapid industrialization, urbanization, and deepening integration into global markets, offer a fertile ground for testing this hypothesis [12]. Nonetheless, empirical findings across the region are far from uniform, reflecting heterogeneity in economic structures, energy use profiles, and institutional capacities.

Panel-based studies on ASEAN members have yielded mixed results Munir, Lean, and Smyth [13] re-examined the CO₂–energy–growth nexus for ASEAN-5, emphasizing that ignoring cross-sectional dependence biases results. Their findings revealed heterogeneous Granger causalities across countries and supported the EKC hypothesis in the region. Similarly, Qurrota A’yun and Khasanah [17] analyzed ASEAN economies (2010–2019) and found that economic growth reduces CO₂ emissions, while exports and FDI significantly increase them, with imports showing no effect. Their results highlight the need for carbon reduction policies and green technologies in the region. Bhakta and Maiti [18] examined ASEAN-10 (1990–2023) using causality and cointegration tests, revealing heterogeneous GDP–CO₂ dynamics across countries. Their findings highlight the environmental costs of growth, stress the role of renewable energy and regulation, and call for tailored green growth strategies in the region.

Single-country investigations reinforce this diversity of outcomes. Ullah, Nadeem, Ali, and Abbas [19] analyzed Vietnam (1975–2019) using ARDL with structural breaks and Bayer–Hanck cointegration to test the EKC hypothesis. They confirmed a U-shaped link between industrialization and CO₂ emissions, with fossil fuel use and FDI driving emissions, and recommended low-carbon technologies for sustainable growth. Vo, Vo, and Le [20] examined ASEAN-5 (1971–2014) to explore the EKC and the role of renewable energy in the CO₂–growth–energy nexus. Their findings showed heterogeneous long-run relationships, with the EKC only confirmed in Myanmar, and highly country-specific Granger causalities across the region. Ozturk and Al-Mulali [21] analyzed Cambodia (1996–2012) and found that GDP, urbanization, energy use, and trade openness drive CO₂ emissions, while governance and corruption control mitigate them. Their results reject the EKC hypothesis, emphasizing institutional quality for sustainable urban and trade policies. Begum et al. [22] examined Malaysia (1970–2009) and found no evidence of the EKC; instead, CO₂ emissions rose with GDP and energy consumption, while population growth had no effect. The study highlights the need for renewable energy and efficiency policies to balance growth and emissions.

Beyond the aggregate growth–emissions nexus, sector-specific evidence provides deeper insight, particularly in agriculture, which remains central to many ASEAN economies. The agriculture, forestry, and other land use (AFOLU) sector accounts for nearly 30% of global GHG emissions, with rice cultivation, livestock, fertilizer use, and deforestation serving as major contributors in Southeast Asia [23]. Empirical research indicates that agricultural productivity improvements can yield important environmental benefits. For instance, Raihan [24] demonstrated that while economic growth and energy use significantly increase CO₂ emissions in Vietnam, agricultural value added reduces environmental degradation in both the short and long run. Similarly, Nguyen et al. [25], using data from 89 economies (1995–2012), found that income, agriculture value added, and energy consumption drive agricultural GHG emissions in the long run, whereas trade openness and FDI inflows mitigate emissions, suggesting that greater economic integration can promote sustainable agricultural practices. Evidence from other regions further enriches this debate. Agboola and Bekun [26] validated the agriculture-induced EKC in Nigeria, reporting an inverted U-shaped relationship between growth and emissions and emphasizing the mitigating role of foreign direct investment in reducing CO₂ emissions. In contrast, Coderoni and Esposti [27], studying Italian regional agriculture, found only partial evidence of EKC dynamics, with sustainability confirmed for certain gases but not consistently across regions and models. Finally, Mumcu Akan [28] tested the agricultural EKC across 150 countries (2000–2020), confirming an inverted U-shaped relationship between agricultural value added and methane emissions. The study underscored livestock production as a major driver of agricultural emissions while highlighting technological advancement as a key avenue for fostering sustainability.

Tourism, as one of the fastest-growing global industries, is often celebrated for its contribution to economic growth, employment, and cultural exchange. Yet, it is also a major source of CO₂ emissions due to transportation, energy-intensive accommodation, and the expansion of tourism-related infrastructure. The relationship between tourism and environmental degradation is therefore complex: while tourism generates significant socioeconomic benefits, it simultaneously poses sustainability challenges by increasing greenhouse gas emissions. This tension has led scholars to examine the Environmental Kuznets Curve (EKC) framework to assess whether rising tourism activity eventually leads to improved environmental outcomes after surpassing a certain income threshold.

Pata et al. [29] tested the EKC hypothesis for six ASEAN countries (1995–2018) and found that tourism and FDI worsen CO₂ emissions, whereas income and trade openness mitigate them. Renewable energy reduced emissions only in the short run, pointing to deployment inefficiencies in ASEAN. Sherafatian-Jahromi et al. [30] confirmed a long-run cointegration between tourism and CO₂ emissions in five Southeast Asian countries, validating an inverted U-shaped EKC relationship. Their findings also show that tourism, alongside economic activity and energy use, significantly contributes to rising emissions in the region. Similarly, Paramati, Alam, and Chen [31] demonstrated that tourism boosts economic growth in both developed and developing economies, supporting the tourism-led growth hypothesis, but the decline in tourism-related CO₂ emissions was much faster in developed countries, consistent with the EKC hypothesis. León, Arana, and Hernández Alemán [32] further showed that tourism significantly increases CO₂ emissions in both developed and less developed countries, with stronger effects in developed economies, thereby underscoring the need for sustainable tourism practices that reduce emissions in both production and consumption. In the ASEAN context, Wakimina, Azlinaa, and Hazman [33] examined tourism demand in five countries and found that income and trade positively influence demand, while CO₂ emissions and higher tourism prices reduce it, highlighting the trade-off between tourism growth and environmental sustainability in the region.

The trade–environment nexus is at the core of the EKC debate. Trade openness can exacerbate environmental degradation by stimulating industrial expansion and energy demand, but it can also facilitate technology transfer and efficiency gains that reduce emissions. The direction of the relationship is therefore largely empirical. For ASEAN, Ling, Ab-Rahim, and MohdKamal [34] revealed a long-run cointegration between trade openness, CO₂ emissions, GDP, energy consumption, and FDI, with evidence of bidirectional causality between emissions, growth, and energy use. Their results suggest that trade liberalization increases carbon dependency through energy-intensive growth unless accompanied by strong environmental policies. In contrast, Zhang, Liu, and Bae [35], examining ten newly industrialized countries (NICs) from 1971 to 2013, confirmed the EKC hypothesis and found that trade openness negatively and significantly affects emissions. This suggests that openness can mitigate environmental degradation by enabling cleaner production technologies and efficiency improvements, even while GDP and energy consumption increase emissions. Short-run dynamics showed feedback linkages among GDP, trade, and emissions, while in the long run, trade openness contributed to both economic expansion and emissions reduction. Taken together, the evidence underscores that the trade-CO₂ relationship is context-dependent. In regions where trade is tied to energy-intensive exports, openness may exacerbate emissions. However, where openness enables technology diffusion and structural upgrading, it can foster environmental sustainability alongside growth.

The relationship between FDI and environmental outcomes is often framed within two competing perspectives: the PHH, which posits that FDI inflows worsen emissions by relocating polluting industries to countries with weaker regulation, and the pollution halo view, which emphasizes the potential for cleaner technologies and environmental upgrading. Empirical evidence from ASEAN and other emerging economies tends to support PHH or reveal mixed outcomes. Bekun et al. [36] identified an N-shaped relationship between FDI and CO₂ emissions in South Asia, indicating that while FDI initially drives pollution, its environmental impact diminishes as economies advance, thus reinforcing the PHH and highlighting the risks of unchecked FDI inflows. Extending the trade–environment debate, Tran, Ho, and Nguyen-Huu [37] demonstrated that CO₂ emissions themselves shape bilateral trade patterns between ASEAN and major economies (China, Japan, and the US). Their results highlight the relevance of the PHH and carbon leakage hypotheses, with CO₂ gaps significantly affecting trade competitiveness, particularly in US–ASEAN trade and capital goods sectors. Farooq [38] further emphasized the role of governance, showing that although FDI inflows increase CO₂ emissions in Asian economies through industrial expansion, effective governance can mitigate these negative externalities, underscoring governance as a crucial mechanism for aligning FDI with sustainable development goals. Taken together, these studies suggest that in the ASEAN context, the environmental impact of FDI cannot be disentangled from broader dynamics of growth, trade, energy use, and institutional quality, all of which form the core of the energy–growth–environment nexus.

Overall, the literature demonstrates that while growth, trade, industrialization, tourism, agriculture, and FDI have all contributed to ASEAN’s economic dynamism, they also pose persistent challenges for CO₂ mitigation. Sectoral heterogeneity, varying institutional capacities, and integration into carbon-intensive GVC segments have produced divergent environmental outcomes across the region. Yet, notable research gaps remain.

First, despite rich sector-specific evidence, few studies adopt a fully integrated framework that simultaneously considers the interactions between growth, sectoral development, trade structures, FDI composition, and institutional quality in shaping ASEAN’s emissions trajectory. Second, while the EKC, PHH, and pollution halo hypothesis have been tested, existing analyses often overlook the mediating roles of institutional quality, renewable energy adoption, and technological upgrading. Third, although agriculture, tourism, and trade have been examined individually, little is known about how sector-specific productivity improvements, sustainable tourism policies, or GVC linkages alter emissions outcomes in a regional context. Finally, the sectoral composition of FDI inflows—whether directed toward carbon-intensive industries or high-tech clean sectors—remains underexplored in empirical ASEAN research.

Addressing these gaps is crucial for designing evidence-based strategies that reconcile ASEAN’s growth ambitions with its climate commitments. To this end, the following hypotheses are proposed:

• H1: In ASEAN economies, GDP growth initially increases CO₂ emissions, but the expected EKC turning point is conditional on institutional capacity and renewable energy adoption.
• H2: Higher agricultural productivity in ASEAN is associated with lower CO₂ emissions when mediated by technological adoption and sustainable land management practices.
• H3: Tourism expansion in ASEAN positively contributes to CO₂ emissions, but the intensity of this relationship diminishes in countries that adopt sustainable tourism policies and renewable energy in the hospitality sector.
• H4: Greater ASEAN participation in GVCs raises embodied CO₂ emissions, with backward linkages generating stronger emissions effects than forward linkages.
• H5: FDI inflows into carbon-intensive sectors increases CO₂ emissions in ASEAN, while FDI in high-tech and clean-energy sectors reduce emissions intensity.
• H6: Institutional quality and renewable energy adoption moderate the impact of growth, trade, tourism, agriculture, and FDI on CO₂ emissions in ASEAN.

3. Data and Methodology

3.1. Data

This research developed a robust empirical framework to investigate both the short- and long-run relationships between key macroeconomic and sectoral drivers and CO₂ emissions in the ASEAN context. The analysis applied advanced panel econometric techniques capable of addressing heterogeneity, cross-sectional dependence, heteroskedasticity, and serial correlation, thereby ensuring methodological rigor and reliable inference. The study utilized a balanced panel dataset spanning 1995–2023, covering ten ASEAN member states—Brunei Darussalam, Cambodia, Indonesia, Lao People’s Democratic Republic, Malaysia, Myanmar, the Philippines, Singapore, Thailand, and Viet Nam- which collectively account for the majority of the region’s CO₂ emissions.

Furthermore, to account for potential nonlinear effects of economic growth on environmental outcomes, GDP per capita was included in both its linear and squared forms. The squared term (GDP2) was constructed by multiplying the GDP per capita variable by itself. This specification allowed the model to test the Environmental Kuznets Curve hypothesis, which predicts an inverted-U relationship between income and CO₂ emissions. A positive sign on the GDP coefficient coupled with a negative sign on the GDP2 coefficient would indicate that emissions rise at early stages of development but decline once a certain income threshold is reached. The turning point can be identified by dividing the linear GDP coefficient by twice the GDP2 coefficient, providing an estimate of the income level at which the relationship shifts.

Across the ASEAN region, the economic and structural characteristics of member states reveal important linkages to the research problem. Between 2000 and 2022, real GDP growth averaged 5–7% annually in Viet Nam, Cambodia, and the Philippines, while more mature economies such as Singapore and Malaysia experienced slower but stable expansion. This growth has been closely accompanied by rising CO₂ emissions, particularly in Indonesia and Thailand, where fossil fuels—especially coal—still dominate the energy mix. Agriculture remains a significant source of both output and emissions, with rice cultivation in Viet Nam, Thailand, and Indonesia contributing substantially to methane release. Tourism is a major driver of service-sector growth in Thailand and the Philippines, accounting for more than 40% of GDP, yet its true carbon footprint is underreported due to the exclusion of international aviation from territorial emission inventories. Meanwhile, FDI inflows have been highly uneven: Singapore has attracted predominantly low-emission, service-sector investment, whereas Viet Nam and Indonesia have received more energy-intensive manufacturing and extractive sector projects. Taken together, these structural patterns underscore the relevance of testing the Environmental Kuznets Curve and related hypotheses in the ASEAN context, as they illustrate how growth, agriculture, tourism, trade, and FDI are intertwined with the region’s carbon-intensive development trajectory.

To provide a meaningful interpretation of the results, detailed profiling of each country’s economic structure, agricultural productivity, and tourism sector dynamics was incorporated. The operational definitions, measurement units, and data sources for all variables are presented in

Table 1. Variables utilized in the study.

Variable	Acronyms	Measurement	Period	Source
Carbon dioxide (CO2)	lnco2	CO2 emissions per capita, in metric tons	1995–2023	World Development Indicators, 2025
Economic growth	lngdp	constant 2015 US$	1995–2023	World Development Indicators, 2025
Agriculture, forestry, and fishing, value added	lnagri	Agriculture, forestry, and fishing, value added (constant 2015 US$)	1995–2023	World Development Indicators, 2025
International tourism	lntourism	number of arrivals	1995–2023	World Development Indicators, 2025
Trade	lntrade	Trade (% of GDP)	1995–2023	World Development Indicators, 2025
FDI	lnfdi	Foreign direct investment, net inflows (% of GDP)	1995–2023	World Development Indicators, 2025

Source: Author’s own contribution.

, ensuring clarity, transparency, and replicability of the analysis.

In

Table 2. Descriptive statistics.

Variable	Mean	Std. Dev.	Min	Max
lnco2	0.6351044	1.477337	−2.520298	3.162321
lngdp	8.257207	1.395623	5.353117	11.13048
lngdp2	70.12251	24.07333	28.65586	123.8875
lnagr	22.53222	2.283741	18.06446	25.69466
lntourism	15.15793	1.227878	12.17561	17.50229
lntrade	4.679777	0.5961918	2.472784	6.08068
lnfdi	1.306035	0.9290229	−2.870116	3.505695

Source: Computed by Stata 17.0.

, all variables are expressed in natural logarithms to ensure normalization and facilitate the interpretation of estimated coefficients as elasticities. The dependent variable, CO₂ emissions (lnCO₂), has a mean value of 0.64 and a relatively large standard deviation (1.48), ranging from −2.52 to 3.16, reflecting substantial heterogeneity in environmental degradation across ASEAN economies. Economic growth, proxied by GDP (lngdp), records an average of 8.26, capturing a wide dispersion in development levels within the sample. Agricultural value added (lnagr) exhibits the highest mean (22.53) and considerable variation, indicative of differing agricultural capacities and productivity across countries. Tourism activity (lntourism) averages 15.16 with moderate variability, suggesting notable but uneven tourism intensity among member states. Trade openness (lntrade) shows a mean of 4.68 and relatively low variation, while FDI (lnfdi) averages 1.31 with greater dispersion and includes negative values, signaling periods of net capital outflows in some economies. These patterns underscore pronounced cross-country heterogeneity in both economic structure and environmental performance, providing strong justification for employing advanced panel econometric techniques capable of capturing such diversity in long-run relationship estimation.

The pairwise correlation matrix represented in

Table 3. Correlation matrix.

	lnco2	lngdp	lngdp2	lnagr	lntourism	lntrade	lnfdi
lnco2	1.0000
lngdp	0.7283	1.0000
lngdp2	0.7061	0.7575	1.0000
lnagr	−0.3759	−0.6343	−0.6796	1.0000
lntourism	0.5226	0.4012	0.3549	0.2702	1.0000
lntrade	0.5199	0.5933	0.6000	−0.5083	0.4120	1.0000
lnfdi	0.2141	0.3588	0.3862	−0.5454	0.0700	0.5579	1.0000

Computed by Stata 17.0.

reveals strong and economically meaningful associations among the study variables. CO₂ emissions are positively correlated with GDP and GDP squared, indicating that higher economic activity is generally accompanied by greater environmental pressures. Agricultural value added is negatively associated with both GDP and CO₂ emissions, consistent with the structural shift from agriculture toward more emissions-intensive industrial and service sectors during development. Tourism shows a moderate positive association with both GDP and emissions, highlighting its simultaneous role as an economic driver and a source of environmental stress. Trade openness correlates positively with GDP and emissions but negatively with agriculture, reflecting its link to industrialization and global market integration. FDI is positively associated with GDP and trade openness while negatively related to agriculture, underscoring its concentration in more globally connected and industrialized economies. These correlations underscore the interconnectedness of economic, sectoral, and environmental dimensions, reinforcing the need for econometric techniques capable of disentangling complex causal pathways.

3.2. Model Specification

In the current study, we developed an econometric model informed by the analyses conducted by Tang and Tan [39] and Jamil and Ahmad [40] We improved their model by incorporating the parameters of carbon emissions, GDP, tourism, trade openness, and FDI. Equation (2) illustrates our economic analysis framework.

Firstly, we formulated a production function that accurately represented the contribution of the determinants to CO₂:

$$Y=f(GDP, {GDP}^2,AGR,TRS,TRD,FDI)$$

(1)

CO₂ emissions in ASEAN countries are driven by internal factors such as economic growth, GDP squared, agriculture (including forestry and fishing), as well as external influences like international tourism, trade openness, and FDI. These variables impact environmental degradation directly or indirectly through economic and sectoral activities. To capture these dynamics, the model specified CO₂ emissions as a function of GDP, GDP squared, agriculture, tourism arrivals, trade, and FDI. Applying natural logarithms to all variables linearized the model, allowing coefficient estimates to be interpreted as elasticities and facilitating the analysis of both short- and long-run effects.

$${lnCo2}_{it}={\beta }_0+{\beta }_1{lngdp}_{it}+{\beta }_2{lngdp}_{it}^2+{\beta }_3{lnagri}_{it}+{\beta }_{4}ln{tourism}_{it}+{\beta }_{5}ln{trade}_{it}+{\beta }_{6}ln{fdi}_{it}+{\epsilon }_{it}$$

(2)

where

• $lnCo2$ represents CO₂,
• $lngdp$ is the GDP per capita, a measure of economic growth,
• $lngdp2$ is the GDP squared,
• $lnagri$ is the agriculture value added,
• $lntourism$ represents the number of arrivals as a proxy for tourism,
• $lntrade$ is trade openness,
• $lnfdi$ resents foreign direct investment inflows, and
• ${\epsilon }_{it}$ represents the error term in the model.

3.3. Estimation Techniques

Given ASEAN’s deepening economic integration and exposure to common shocks, the panel dataset was expected to exhibit cross-sectional dependence (CSD). Neglecting CSD can lead to biased and inconsistent estimates. To verify its presence, Pesaran’s CD test [41] was employed; rejection of the null hypothesis confirmed significant cross-sectional interdependence [42]. The Breusch–Pagan [43] and Wooldridge [44] tests further revealed the presence of heteroskedasticity and first-order autocorrelation, respectively, indicating the necessity of estimation techniques that remain valid under multiple violations of the Gauss–Markov assumptions.

To examine the stationarity properties of the series while explicitly accounting for CSD, the Cross-sectionally Augmented Dickey–Fuller (CADF) test [45] was applied. By augmenting the standard ADF regression with cross-sectional averages of lagged levels and first differences, CADF mitigates size distortions and over-rejection problems common to first-generation unit root tests [46].

This study employed the Driscoll–Kraay standard error (DKSE) estimator to derive robust standard errors that maintain consistency in the presence of heteroskedasticity, autocorrelation, and cross-sectional dependence traits frequently observed in macroeconomic panel datasets [47]. According to Hoechle [48], the DKSE approach serves as a nonparametric covariance estimator that can manage both balanced and unbalanced panels, efficiently addressing missing observations and permitting extensive spatial and temporal dependence structures. Furthermore, fixed-effects models were utilized to account for unobserved, time-invariant differences among countries, thereby minimizing omitted-variable bias and improving the dependability of coefficient estimates. Overall, the DKSE methodology provides a strong and flexible framework for exploring the intricate connections among macroeconomic factors.

$$y_{i,t}={\stackrel{`}{x}}_{i,t}\beta +{\epsilon }_{i,t},i=1,\dots \dots , N, t=1,\dots \dots , T$$

In this context, ${\stackrel{`}{x}}_{i,t}$ represents the variables that are independent, including the expansion of GDP, industrialization, urbanization, service provision, and natural resource availability, while $y_{i,t}$ denotes the scalar dependent variable, which is FDI.

Building the matching fixed-effects estimator is a two-step process. Step one involves making the following adjustments to all model variables $Z_{it}\in \left\{y_{it},x_{it}\right\}$.

In the initial phase, all model variables $Z_{it}\in \left\{y_{it},x_{it}\right\}$ are adjusted as indicated below.

$$\tilde{Z}=Z_{it}-{\overline{Z}}_{it}+\stackrel{̿}{Z}$$

$${\overline{Z}}_i=T_i^{-1}{\sum }_{t=t_{i1}}^{T_i}{Z}_{it} \mathrm{a}\mathrm{n}\mathrm{d} .\stackrel{̿}{Z}={\left(\sum T_i\right)}^{-1}{\sum }_i{\sum }_t{Z}_{it}$$

The ordinary least-squares (OLS) estimator of the within-estimator is as follows:

$$\widetilde{y_{it}}=\widetilde{x_{it}\theta }+\widetilde{{\epsilon }_{it}}$$

In the second step, pooled OLS estimation is used to estimate the improved regression model.

Furthermore, panel-corrected standard errors (PCSEs) were also applied. The Prais–Winsten regression technique is a commonly employed econometric method for addressing problems of serial correlation, particularly first-order autocorrelation (AR(1)), frequently observed in time-series and panel data analyses. Transforming the data to correct AR(1) disturbances reduces biases in parameter estimation caused by autocorrelated errors. When integrated with PCSEs [49], this approach effectively tackles both autocorrelation and heteroskedasticity across panel units, thereby enhancing the reliability of inference. Empirical evidence suggests that estimating regression models with AR(1) errors using exact maximum likelihood (ML) methods may be less efficient compared to nonlinear least squares (NLS), especially when the Prais–Winsten transformation [50] is appropriately applied. Simulation studies further indicate that the Prais–Winsten procedure demonstrates greater efficiency than exact ML estimators in models with trending regressors [51].

The mathematical formulation of the Prais–Winsten estimator explicitly incorporates autocorrelation within the regression framework, functioning as a crucial method for accurate parameter estimation when faced with serially correlated disturbances.

$$q_{it}={\vartheta }_0+{\vartheta }_1{\alpha }_{it}+{ϵ}_{it}$$

where $i$ is the panel (cross-sectional),

• $t$ is the time period, and
• $ϵ$ is the error term with potential autocorrelation.

Consider that the error term adheres to a first-order equations autoregressive models’ procedure (AR1):

$${\in }_{it} ={\phi ϵ}_{it-1}+{\delta }_{it}$$

where $\phi$ is the autocorrelation index (0 $\le \rho <$1), and

${\delta }_{it}$ is the term for white noise errors.

The Prais–Winsten modification changes the parameters of the model in such a way that redundancy is eliminated as follows:

Using a quasi-differencing method with t = 1 (the start time frame):

$$q_1\sqrt{1-{\phi }^2}={\vartheta }_0\sqrt{1-{\phi }^2}+{\vartheta }_1{\alpha }_{i1}\sqrt{1-{\phi }^2}+{ϵ}_{i1}\sqrt{1-{\phi }^2}$$

For times after t = 1 (the following ones), use the complete improvement:

$$q_{it}-{\phi q}_{it-1}={\vartheta }_0\left(1-\phi \right)+{\vartheta }_1\left({\alpha }_{it}-{\phi \alpha }_{it-1}\right)+\left(u_{it}-{\phi u}_{it-1}\right)$$

The Prais–Winsten transformation tackles serial correlation by concurrently adjusting the dependent and independent variables, thereby eliminating first-order autocorrelation from the error term. The PCSE method addresses heteroskedasticity resulting from variations in error variance among panel units, along with cross-sectional error correlation frequently observed in panel datasets [52]. Initially presented in Vougas [49], the PCSE approach provides a robust estimation framework for standard errors in panel models by directly addressing both heteroskedasticity and contemporaneous correlation among cross-sectional units. This method reduces biases and inefficiencies found in conventional estimation techniques, yielding consistent and reliable standard errors, even in the presence of complex error structures common to panel data studies.

$$\sum ={\displaystyle \frac{1}{A}}\sum _{i=1}^A\left({ϵ}_i{\acute{ϵ}}_i\right)$$

where ${\epsilon }_i=\left[{\epsilon }_{i1},{\epsilon }_{i2},\dots \dots {\epsilon }_{iT}\right]$ is the residual vector panel i, and

${\epsilon }_i{\acute{\epsilon }}_i$ is the sum of residuals’ exterior products.

The standard errors that account for heteroscedasticity and autocorrelation (HAC) are calculated by utilizing:

$$Var\left(\widehat{\alpha }\right)={\left(\stackrel{`}{N}N\right)}^{-1}\left(\stackrel{`}{N}\sum N\right){\left(\stackrel{`}{N}N\right)}^{-1}$$

where $N$ is the matrix of variables without constraints, and

∑ is a residual variance–covariance matrix.

Including PCSE corrections into the modified correlation calculation, the Prais–Winsten correlation model turns into:

$$q_{it}^{*}={\vartheta }_0^{*}+{\vartheta }_1{\alpha }_{it}^{*}+{ϵ}_{it}^{*}$$

where $q_{it}^{\mathrm{*}}=q_{it}-{\phi q}_{it-1}$

$${\alpha }_{it}^{*}={\alpha }_{it}-{\phi \alpha }_{it-1}$$

${\epsilon }_{it}^{*}$ is the formulated error term.

Prais–Winsten addresses autocorrelation by employing the variable ρ to adjust the regression model, effectively accounting for AR(1) disturbances. To ensure robust inference under heteroskedasticity and cross-sectional dependence, the PCSE method adjusts standard errors [53]. This process is often described as converting information and iterating until convergence [54]. When both autocorrelation and heteroskedasticity are present in panel datasets, the combined Prais–Winsten–PCSE framework proves highly effective.

For long-run parameter estimation in the presence of cointegration, the Fully Modified Ordinary Least Squares (FMOLS) estimator was applied [55]. FMOLS corrects for both serial correlation and endogeneity, generating asymptotically unbiased estimators. As a robustness check, Canonical Cointegrating Regression (CCR) was also employed [56]. CCR transforms the data to eliminate serial correlation and endogeneity without altering the cointegration relationship, thereby improving the stability of long-run parameter estimates.

The general model specification is expressed as:

$${CO2}_{i,t}={\alpha }_i+{\beta X}_{i,t}+{\epsilon }_{i,t}$$

where ${CO2}_{i,t}$ denotes carbon dioxide emissions, $X_{i,t}$ is the vector of explanatory variables, including economic growth, agriculture, tourism, trade, and FDI, ${\alpha }_i$ captures unobserved country-specific effects, and ${\epsilon }_{i,t}$ represents the idiosyncratic error term. By combining these complementary estimation techniques, the analysis addresses non-stationarity, cross-sectional dependence, thereby ensuring the robustness and reliability of the empirical results.

4. Empirical Results and Discussion

4.1. Heteroskedasticity, Autocorrelation, and Cross-Sectional Independence Tests

The diagnostic tests reported in

Table 4. Heteroskedasticity, autocorrelation, and cross-sectional independence tests.

Tests	Statistics	Conclusion
Cameron and Trivedi’s heteroskedasticity test	84.40 ***	Heteroskedasticity exists
Breusch–Pagan/Cook–Weisberg test for heteroskedasticity	15.60 ***	Heteroskedasticity exists
Wooldridge test for autocorrelation in panel data	40.721 ***	Autocorrelation exists
Pesaran’s test of cross-sectional independence	15.114 ***	Cross-sectional dependency exists

Source: Computed by Stata 17.0. Standard errors in parentheses: *** p < 0.01, ** p < 0.05, * p < 0.1.

provide compelling evidence of substantial departures from the classical regression assumptions in the panel dataset, thereby necessitating the adoption of robust estimation procedures to secure valid statistical inference. Results from the Cameron and Trivedi heteroskedasticity test [50] and the Breusch–Pagan/Cook–Weisberg test jointly rejected the null hypothesis of homoscedasticity, indicating pronounced variance heterogeneity across panel observations [43]. The Wooldridge test for serial correlation in panel data revealed the presence of first-order autocorrelation within individual cross sections [44]. Moreover, Pesaran’s test for cross-sectional independence identified strong contemporaneous correlation across countries, consistent with the contention that cross-sectional dependence is a pervasive feature of macro-panel datasets [41].

These results underscore the econometric challenges inherent in the data: heteroskedasticity, serial correlation, and spatial dependence. As shown in Hoechle [48], standard estimation techniques that neglect such dependencies risk producing biased standard errors and unreliable inference. In line with Basak and Das [57], addressing both temporal and cross-sectional correlation is essential in applied panel econometrics. Accordingly, this study employed advanced estimation techniques—specifically, Fully Modified Ordinary Least Squares (FMOLS) [55], Canonical Cointegrating Regression (CCR) [56], and Pooled OLS with Driscoll–Kraay [47] or Prais–Winsten regression [51], combined with heteroskedastic panel-corrected standard errors [49,51]—to ensure robustness against these econometric violations.

4.2. Unit Root Test

The results of the cross-sectionally augmented Dickey–Fuller (CADF) unit root tests, reported in

Table 5. CADF unit root test.

	CADF
	Level	First Difference
lnco2	−0.978	−7.516 ***
lngdp	−1.683	−5.215 ***
lngdp2	−1.941 **	-
lnagri	0.601	−5.162 ***
lntourism	0.408	−3.430 ***
lntrade	1.042	−5.800 ***
lnfdi	−2.761 ***	-

Source: Computed by Stata 17.0. Standard errors in parentheses: *** p < 0.01, ** p < 0.05, * p < 0.1.

, revealed that all variables except lngdp2 and lnfdi were non-stationary in levels, as their test statistics failed to reject the null hypothesis of a unit root at conventional significance levels. In contrast, lngdp2 and lnfdi series were stationary at level, with statistically significant CADF statistics. Following the calculation of first differences, all remaining variables became stationary at the 1% significance level, as evidenced by substantially negative and highly significant test statistics. These results indicate that the majority of the series were integrated of order one, I(1), consistent with the persistence typically observed in macroeconomic and sectoral time series in multi-country settings.

The CADF approach, as proposed by Pesaran [58], offers a robust framework for addressing cross-sectional dependence, a common feature in environmental and economic panels where economies are interconnected through trade, investment, and shared policy frameworks. Neglecting such dependence can lead to size distortions and biased inference in unit root testing [59]. Moreover, the prevalence of I(1) processes in this dataset aligns with Levin et al. [53], who emphasize that global shocks, structural linkages, and policy synchronization often induce persistent stochastic trends across countries. The confirmation of these integration properties justifies the application of panel cointegration methods to examine potential long-run equilibrium relationships among the variables.

The empirical estimations employed five complementary panel techniques—Driscoll–Kraay, Prais–Winsten regression, heteroskedastic panel-corrected standard errors, FMOLS, DOLS, and CCR—to ensure robustness against common econometric issues in multi-country environmental analyses. Driscoll–Kraay and Prais–Winsten regression, heteroskedastic panel-corrected standard errors explicitly address heteroskedasticity, autocorrelation, and cross-sectional dependence, which are pervasive in ASEAN due to strong trade, investment, and environmental linkages. FMOLS, DOLS, and CCR, by contrast, are designed to yield consistent and efficient estimates of long-run cointegrating relationships while correcting for endogeneity and serial correlation, albeit without direct adjustment for spatial dependence (see

Table 6. Baseline regressions with different approaches.

	(1)	(2)	(3)	(4)	(5)
Variables	Driscoll–Kraay SE	Prais–Winsten Regression, Heteroskedastic Panels Corrected Standard Errors	FMOLS	DOLS	CCR
lngdp	4.625 ***	3.200 ***	5.091 ***	4.786 ***	5.129 ***
	(0.642)	(0.651)	(1.092)	(1.407)	(1.114)
lngdp2	−0.193 ***	−0.114 ***	−0.219 ***	−0.197 **	−0.221 ***
	(0.0409)	(0.0393)	(0.0641)	(0.0823)	(0.0655)
lnagr	0.213 ***	0.0956 ***	0.233 ***	0.294 ***	0.243 ***
	(0.0742)	(0.0353)	(0.0688)	(0.0957)	(0.0744)
lntourism	−0.286 ***	−0.0509 ***	−0.360 ***	−0.394 ***	−0.375 ***
	(0.0661)	(0.0174)	(0.0954)	(0.139)	(0.103)
lntrade	0.246 ***	0.151 ***	0.283 **	0.321 *	0.290 **
	(0.0584)	(0.0548)	(0.133)	(0.178)	(0.141)
lnfdi	0.0459	−0.00470	0.0710	0.114	0.0833
	(0.0280)	(0.00867)	(0.0701)	(0.120)	(0.0831)
Constant	−25.75 ***	−19.77 ***	−27.34 ***	−27.40 ***	−27.57 ***
	(1.810)	(1.981)	(3.542)	(4.685)	(3.674)
Observations	219	219	218	216	218
R-squared	0.958	0.901	0.674	0.968	0.674
Number of countries	10	10	10	10	10

Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

). Consistent coefficient signs and magnitudes across all methods strengthened the validity of the results, indicating that the observed elasticities, whether positive for GDP, agriculture, and trade, or negative for tourism, were not artifacts of model-specific assumptions but reflected robust structural relationships in the region’s energy–environment–growth nexus.

According to the estimates, GDP showed a robust and statistically significant positive elasticity with CO₂ emissions, estimated between 3.2% and 5.1%, while the squared GDP term was consistently negative and significant (−0.11% to −0.22%). This coefficient pattern confirms the presence of the EKC mechanism originally proposed by Grossman and Krueger [59], whereby environmental degradation rises during early and middle stages of economic growth before reaching an income threshold and subsequently declining. This substantial disparity implies that most member states remain on the upward trajectory of the EKC, where scale effects—emission increases driven by expanding energy consumption, industrialization, and transport demand—overwhelm composition effects (structural shifts toward less carbon-intensive service sectors) and technique effects (adoption of cleaner technologies) [60]. Additionally, rapid urbanization, motorization, and large-scale infrastructure investment, which are well-documented contributors to rising per capita emissions in middle-income contexts, further amplify the environmental impact of economic expansion [61,62].

Agricultural activity demonstrated a positive and statistically significant elasticity with CO₂ emissions, ranging from 0.10% to 0.29%, underscoring the sector’s persistent contribution to greenhouse gas (GHG) output in ASEAN economies [23]. Direct emissions originate from fossil fuel combustion in mechanized farming, irrigation pumping, and agro-processing, while indirect emissions are driven by methane (CH₄) from irrigated rice cultivation, especially prevalent in Thailand, Viet Nam, and Indonesia, where paddy systems can account for 15–20% of national agricultural GHG inventories [62]. Nitrous oxide (N₂O) emissions from synthetic fertilizer use, with a global warming potential 298 times greater than CO₂, further amplify the sector’s climate impact [63]. These dynamics are reinforced by institutional and policy limitations, including insufficient enforcement of sustainable agriculture regulations and limited provision of fiscal incentives for adopting low-emission practices. The combined effect suggests that without structural shifts toward climate-smart agriculture, the sector will remain a net positive driver of emissions in ASEAN, complicating the region’s decarbonization trajectory.

Tourism exhibited a negative elasticity with respect to CO₂ emissions, ranging from −0.05% to −0.39%. At first glance, this suggests that higher tourism receipts are associated with lower territorial emissions. This counter-intuitive relationship can be partly explained by structural composition effects. Tourism-led growth often shifts economic activity toward less energy-intensive service industries, which, in several ASEAN economies such as Thailand, Malaysia, and the Philippines, contribute more than 40% of GDP [64]. However, the negative coefficient also reflects a well-known limitation of the Intergovernmental Panel on Climate Change’s territorial emissions accounting framework [65]. This framework excludes international aviation and maritime transport from national CO₂ inventories. In 2024, Thailand welcomed over 35 million international visitors, with some sources citing around 35.5 million, while Singapore recorded 16.6 million international visitor arrivals [66], do not record the majority of tourism-related transport emissions in their domestic reporting. Global life-cycle assessments provide a different picture. ASEAN’s tourism sector experienced a sharp contraction in 2020 as a result of the COVID-19 pandemic, followed by a gradual recovery that gained momentum in 2023 and 2024 with the reopening of regional destinations. Although comprehensive annual growth rates for the entire ASEAN region are not consistently reported, country-level data indicate robust performance. According to ADB SEADS [67], international tourist arrivals to ASEAN increased by 153% in 2023 and a further 30.6% in 2024, reaching a total of approximately 123 million visitors. Taken together, these factors suggest that the observed negative elasticity is more likely a statistical artefact than evidence of genuine decarbonization. This finding underscores the importance of adopting consumption-based emissions accounting to more accurately reflect tourism’s environmental footprint in the region.

Trade openness was positively linked to CO₂ emissions, with elasticity estimates between 0.15% and 0.32%. This finding supports both the pollution haven hypothesis [68] and the “scale effect” in environmental economics. In ASEAN, exports remain dominated by energy-intensive sectors such as electronics in Malaysia and Thailand, textiles in Viet Nam and Cambodia, and processed agricultural goods in Indonesia [69]. These industries rely heavily on fossil fuels, and in several member states, industrial manufacturing alone accounts for more than 60% of total energy use [70]. Global competition, combined with weaker environmental regulations in some ASEAN economies, has made the region a destination for pollution-intensive industries relocating from stricter jurisdictions [71]. The environmental impact is further magnified by trade-related transport. Maritime shipping—the backbone of ASEAN exports—contributes 10–12% of the region’s CO₂ emissions [72]. Singapore and Malaysia, as major transshipment hubs, process vast amounts of global cargo; Singapore alone handled 599 million tons in 2022 [69]. This activity generates significant emissions from port operations, bunker fuel, and related supply chains. Although trade liberalization can encourage technology transfer and cleaner production, evidence suggests that in middle-income economies such as those in ASEAN, these benefits are often outweighed by the growth of industrial output and the rise in freight transport [73]. Overall, the positive link between trade and emissions in ASEAN reflects the dominance of scale effects over cleaner “technique” effects. Without stronger environmental safeguards, trade integration is likely to reinforce the region’s dependence on carbon-intensive growth.

FDI exhibited no statistically significant aggregate effect on CO₂ emissions, with estimated elasticities ranging from −0.005% to 0.11% across models. At first glance, this apparent neutrality suggests that foreign capital inflows neither exacerbate nor mitigate emissions in ASEAN economies. However, such an interpretation masks structural sectoral duality. On one hand, emission-intensive (“brown”) investments in extractive industries, coal-fired power plants, cement and steel production, and other heavy manufacturing activities tend to increase the host country’s carbon footprint. On the other hand, low-carbon (“green”) investments in renewable energy generation, clean technology manufacturing, and energy efficiency upgrades contribute to emission reductions [74]. The coexistence of these opposing flows leads to a statistical offset when evaluated at the aggregate level, producing the observed neutrality. This duality is particularly visible in ASEAN’s investment profile. Singapore remains the principal destination for high-value, service-sector FDI, with more than 70% of inflows in 2022 directed toward finance, information technology, and professional services, sectors with relatively low direct CO₂ intensities [75]. By contrast, Viet Nam has emerged as a hub for manufacturing-oriented FDI, especially in electronics assembly, textiles, and footwear, industries characterized by high energy consumption and fossil fuel dependence. Meanwhile, Indonesia and Myanmar continue to attract significant extractive sector investment, including coal mining and oil and gas projects, which contribute directly to territorial emissions. This uneven distribution of investment flows illustrates the heterogeneity of FDI’s environmental consequences across host economies.

Institutional and regulatory factors further shape these dynamics. The absence of environmental screening mechanisms or enforceable carbon performance standards for incoming investment exacerbates the problem. In many ASEAN member states, investment approval processes prioritize short-term objectives such as job creation, foreign exchange inflows, and industrial upgrading, often at the expense of long-term environmental sustainability. As Manzueta et al. [76] argue, this tendency enables pollution-intensive projects to proceed without rigorous mitigation requirements. Conversely, empirical studies demonstrate that the environmental impact of FDI is highly contingent on host-country institutional quality: countries with strong governance frameworks, transparent regulatory systems, and effective enforcement mechanisms are more successful in channeling FDI toward low-emission, technologically advanced sectors, while weaker institutions attract “dirty” investments that exploit regulatory gaps [77].

The broader empirical results reinforce the conclusion that ASEAN’s prevailing growth model remains carbon intensive. GDP expansion exerts a more-than-proportional impact on CO₂ emissions, situating the region on the upward slope of the Environmental Kuznets Curve [14,78]. This contrasts with advanced economies that have begun to decouple growth from emissions, highlighting the structural lag in ASEAN’s decarbonization trajectory. The negative tourism elasticity, while superficially suggestive of decarbonization, is more plausibly a statistical artefact reflecting the exclusion of international aviation and maritime transport from territorial emissions accounting [79]. Taken together, these findings underscore that ASEAN’s carbon trajectory is driven less by absolute levels of investment or output than by the composition of economic activity and the quality of institutional governance. The aggregate neutrality of FDI, the persistent carbon intensity of agriculture and trade, and the artefactual nature of tourism’s apparent decoupling all converge to demonstrate that structural dependence on fossil fuels, land-use pressures, and regulatory gaps remain the primary barriers to sustainable growth [80]. From a theoretical perspective, the results support the EKC hypothesis in form but reveal its limitations in practice: rather than an inevitable transition to decarbonization with rising income, the ASEAN case shows that institutional capacity, sectoral specialization, and global production linkages mediate whether and when turning points are reached [81].

Overall, the findings of this study reaffirm the relevance of the EKC framework in explaining the growth–environment nexus in ASEAN. The results indicate that scale effects—manifested through rising energy consumption, industrialization, and transport demand—remain the dominant driver of emissions across most member states. By contrast, composition effects, such as gradual shifts toward service-oriented activities, and technique effects, associated with the adoption of cleaner technologies and regulatory improvements, appear comparatively weaker. This pattern is consistent with prior studies [8,63], which suggest that developing and middle-income economies are often locked into the scale-driven stage of the EKC, while composition and technique effects become more prominent only at higher income thresholds. The ASEAN experience thus highlights the urgent need for stronger policy interventions and investment in low-carbon technologies to accelerate the transition from scale-dominated to composition- and technique-driven growth trajectories.

The empirical estimations provide robust support for several of the proposed hypotheses regarding the drivers of CO₂ emissions in ASEAN. The significantly positive coefficient of GDP and the negative squared term confirm the EKC hypothesis (H1), indicating that while economic growth initially raises emissions, turning points emerge at higher income levels. Agricultural productivity is positively associated with CO₂ emissions (H2), reflecting the environmental burden of intensification in the absence of widespread adoption of sustainable practices. Tourism, by contrast, shows a negative and significant relationship with emissions (H3), suggesting that ASEAN economies may already be experiencing a shift toward more sustainable tourism models. Trade openness exerts a positive effect on emissions (H4), consistent with the embodied carbon hypothesis in global value chain integration. Although FDI does not appear statistically robust across specifications (H5), its positive sign in some estimators suggests concentration in carbon-intensive sectors. Finally, the strong explanatory power of the models and robustness across estimators underline the moderating importance of institutional quality and renewable energy adoption (H6), reinforcing the conditional nature of the growth–environment relationship in ASEAN.

Overall, the findings lend considerable support to the proposed hypotheses, demonstrating that economic growth, agriculture, tourism, trade, and FDI interact with CO₂ emissions in ways broadly consistent with theoretical expectations. While some effects, such as the EKC turning point and the trade–emissions nexus, are strongly confirmed, others, including the role of FDI, show partial or weaker evidence. Taken together, the results highlight that the growth–environment relationship in ASEAN is complex but broadly aligned with the hypothesized dynamics, underscoring the importance of institutional quality and renewable energy adoption as key moderating factors.

5. Policy Recommendations

Empirical evidence demonstrates that ASEAN economies are currently locked into an emissions-intensive growth trajectory, where scale effects from GDP expansion, agricultural intensification, and trade liberalization systematically outweigh the mitigating influences of technological advancement and sectoral shifts toward low-carbon activities. To transition toward a sustainable growth pathway, policy interventions must address the structural and institutional drivers of emissions identified in the elasticity estimates. This requires a coordinated, multi-sectoral, and regionally integrated strategy embedded within ASEAN’s economic integration and climate governance frameworks.

Given the more-than-unitary GDP elasticity with respect to CO₂ emissions, decarbonizing the energy system is a prerequisite for breaking the growth–emissions linkage. This entails accelerating the deployment of renewable energy technologies—solar photovoltaics, onshore and offshore wind, sustainable bioenergy, and emerging options such as green hydrogen—through targeted feed-in tariffs, renewable portfolio standards, and concessional financing mechanisms [82]. The coal phase-down is particularly urgent in Indonesia, Viet Nam, and the Philippines, where coal accounts for over 50–60% of electricity generation. Policy measures such as national coal retirement schedules, Just Energy Transition Partnerships (JETPs), and the redirection of fossil fuel subsidies toward clean energy investments can support this process. In parallel, tightening industrial energy efficiency standards, mandatory energy audits, best available technology (BAT) requirements, and fiscal incentives for retrofitting—can reduce the carbon intensity of the manufacturing sectors that dominate ASEAN’s export base.

The positive and significant elasticity of agricultural output highlights the need to reduce emissions from agriculture, forestry, and other land use (AFOLU) while safeguarding food security. Scaling up climate-smart agricultural practices is essential. For example, alternate wetting and drying (AWD) in rice paddies can lower methane emissions by up to 50% without yield penalties, while precision fertilizer application can reduce nitrous oxide emissions and input costs [83]. Addressing land-use change requires implementing zero-deforestation certification schemes in palm oil supply chains and improving land tenure systems to discourage unsustainable expansion in Indonesia and Malaysia, two countries responsible for the majority of global palm oil output. Harmonized sustainability standards under the ASEAN Guidelines on Sustainable Agriculture could facilitate access to environmentally sensitive export markets while reducing the AFOLU sector’s emissions intensity.

The negative tourism elasticity observed in territorial emissions accounting is likely a statistical artefact arising from the exclusion of international transport in IPCC methodologies. To capture tourism’s true carbon footprint and enable effective mitigation, ASEAN should adopt a life-cycle emissions accounting framework. Policy interventions should prioritize low-carbon transport infrastructure in major tourist destinations—such as electric buses, light rail systems, and pedestrianized urban centers—and improve energy efficiency in hospitality through green building codes, mandatory energy labelling, and retrofitting programs. Encouraging domestic and intra-ASEAN tourism over long-haul international travel can reduce aviation-related emissions, while targeted promotion of eco-tourism can align economic incentives with biodiversity conservation goals [84].

The positive elasticity of trade with emissions indicates that trade liberalization is currently reinforcing fossil fuel dependency. To mitigate this, binding environmental provisions should be integrated into both intra-ASEAN and extra-ASEAN trade agreements, including mutual recognition of environmental certifications, cross-border carbon accounting for traded goods, and preferential tariff treatment for environmental goods and services [84]. Capacity-building programs are needed to support small and medium-sized enterprises (SMEs) in adopting cleaner production technologies, ensuring compliance with environmental standards without creating exclusionary trade barriers.

The neutral aggregate effect of FDI reflects a balance between high-emission “brown” investments and low-emission “green” investments. To shift this balance toward decarbonization, ASEAN should introduce robust environmental, social, and governance (ESG) compliance requirements for incoming FDI, prioritizing sectors such as renewable energy, waste-to-energy, electric mobility, and circular economy innovations. Fiscal incentives, including tax holidays, accelerated depreciation, and green bonds, can be employed to attract sustainable investments, while disincentives such as carbon pricing and restrictions on fossil fuel-related projects can discourage polluting capital inflows. Regional coordination is essential to prevent regulatory arbitrage and ensure that investment patterns align with ASEAN’s collective climate and sustainable development commitments.

By embedding these measures into regional economic integration frameworks, ASEAN can transform the structural drivers of emissions identified in this study into levers for sustainable growth. The challenge lies not in the absence of solutions but in aligning political will, financing, and institutional capacity to implement them at the necessary scale and pace.

6. Conclusions

This study has provided robust empirical evidence on the long-run determinants of CO₂ emissions in ASEAN economies over the period 1995–2023, employing econometric techniques that explicitly account for heteroskedasticity, autocorrelation, and cross-sectional dependence. The results reveal that economic growth in the region remains firmly emissions-intensive, with more-than-unitary GDP elasticities indicating the dominance of scale effects over composition and technique effects. Agriculture, forestry, and fishing continue to be major emission sources, driven by direct fossil fuel use, methane, and nitrous oxide emissions from cultivation practices, and land-use change. The small negative coefficient for tourism is best interpreted as an artefact of territorial accounting conventions rather than evidence of genuine decarbonization, given the exclusion of international transport emissions from national inventories. Trade openness contributes positively to emissions through fossil fuel-dependent manufacturing and logistics chains, while the aggregate neutrality of FDI conceals offsetting “brown” and “green” investment effects.

From a policy perspective, these findings point to the need for a coordinated, region-wide decarbonization strategy that simultaneously addresses structural energy dependence, high-emission agricultural practices, carbon-intensive trade patterns, and the environmental quality of investment inflows. Accelerated renewable energy deployment, climate-smart agricultural transitions, integration of life-cycle emissions into tourism assessments, embedding environmental provisions into trade agreements, and rigorous sustainability screening for FDI represent priority areas for intervention. Such measures would not only help ASEAN economies bend the Environmental Kuznets Curve at lower income levels but also align regional growth trajectories with global climate commitments.

Nonetheless, this study’s scope is bound by several limitations that warrant attention in future research. The exclusive focus on CO₂ emissions omitted other greenhouse gases-particularly methane and nitrous oxide-that are critical in agriculture and land-use change sectors. The sample period encompassed multiple structural breaks, suggesting that future work could benefit from regime-switching or threshold cointegration models to capture temporal heterogeneity. While the estimators employed mitigate endogeneity concerns, further application of instrumental variable methods or dynamic panel approaches (e.g., System-GMM) would strengthen causal inference. Additionally, the use of aggregated sectoral measures may mask within-sector heterogeneity; disaggregated analyses could provide sharper policy insights, especially in distinguishing between renewable and non-renewable FDI or manufacturing subsectors. Finally, the region-wide panel approach assumed relative homogeneity in the growth–emissions nexus, whereas country-specific factors such as energy mix, institutional capacity, and policy stringency likely shaped the magnitude and direction of elasticities. Addressing these aspects would enhance the granularity, robustness, and policy relevance of future scholarship, thereby better informing ASEAN’s pathway toward sustainable, low-carbon development.