Determinants of CO₂ Emissions in Hydrocarbon-Dependent Economies: A Multi-Method Panel Analysis of GCC Countries

Abstract

This study investigates the key drivers of CO₂ emissions in Gulf Cooperation Council (GCC) countries, focusing on energy consumption, economic growth, urbanization, trade openness, and institutional quality over the period 1980–2023. The analysis adopts a structured econometric framework in which each technique serves a distinct analytical purpose. Fixed-effects regression with Driscoll–Kraay standard errors is used to estimate contemporaneous relationships while accounting for cross-sectional dependence. LASSO regression is employed as an exploratory variable-selection tool to identify the most relevant predictors of emissions, while Common Correlated Effects Mean Group (CCEMG) and instrumental-variable fixed-effects (IV-FE) estimators are used as robustness checks to account for cross-sectional dependence, slope heterogeneity, and potential endogeneity. The results consistently identify energy consumption as the dominant determinant of CO₂ emissions across all specifications, reflecting the persistent dependence of GCC economies on fossil fuel-intensive energy systems. GDP per capita also exhibits a positive relationship with emissions, suggesting a possible non-linear income–emissions relationship; however, the results do not provide robust support for the conventional EKC hypothesis. Population density is negatively associated with emissions, suggesting potential efficiency gains from urban concentration. In contrast, regulatory quality does not show a statistically robust effect and is excluded from the preferred specification by the LASSO procedure. The impact of trade openness appears model-dependent, indicating that its environmental effects vary according to the balance between scale and technology-transfer effects. Overall, the findings provide policy-relevant insights for balancing economic growth, energy transition, and decarbonization objectives in hydrocarbon-dependent economies.

1. Introduction

The Gulf Cooperation Council (GCC) countries, namely Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and the United Arab Emirates, are among the world’s most carbon-intensive economies, with per capita CO₂ emissions exceeding global averages by three to five times (IEA, 2023). As signatories to the Paris Agreement, these hydrocarbon-dependent states face increasing pressure to reconcile hydrocarbon-dependent growth models with international decarbonization commitments and long-term economic diversification objectives. This tension is particularly acute given that the region accounts for over 30% of global oil production (Energy Institute, 2023) while simultaneously experiencing some of the fastest growth rates in energy demand worldwide (IEA, 2023).

From a theoretical perspective, this dilemma reflects a structural tension between rentier-state economic systems and the transition toward sustainable low-carbon development pathways. In such economies, energy-intensive growth, subsidized fuel prices, and export-oriented hydrocarbon sectors generate persistent carbon lock-in effects, limiting the effectiveness of conventional environmental policies.

Despite increasing scholarly attention to the determinants of environmental degradation, important questions remain regarding the relative importance of economic, demographic, institutional, and energy-related factors in hydrocarbon-dependent economies such as those of the Gulf Cooperation Council (GCC). Existing studies have extensively examined the roles of energy consumption, economic growth, urbanization, trade openness, and institutional quality in explaining carbon emissions. However, the empirical evidence remains mixed and often context-dependent, particularly in resource-rich economies characterized by strong dependence on fossil fuels, extensive energy subsidies, and ongoing economic diversification efforts.

Previous research has frequently incorporated several of these determinants simultaneously, both in GCC countries and in other regions. Nevertheless, differences in data coverage, model specifications, econometric techniques, and institutional contexts have produced conflicting findings regarding the magnitude, direction, and significance of their effects on environmental quality. Consequently, there remains no clear consensus regarding the dominant drivers of emissions in hydrocarbon-dependent economies or the extent to which governance improvements can contribute to environmental sustainability.

The role of institutional quality is particularly controversial. A substantial body of literature suggests that stronger governance, regulatory effectiveness, and institutional quality can improve environmental outcomes by strengthening policy implementation and environmental enforcement (Danish et al., 2020; Hassan et al., 2018). However, other studies report weak, insignificant, or context-specific effects, especially in economies where fossil-fuel dependence and economic growth objectives may constrain environmental policy effectiveness (Alkasasbeh et al., 2023). These mixed findings highlight the need for additional evidence from GCC countries.

Accordingly, the contribution of this study does not lie in the inclusion of specific explanatory variables, which have been widely examined in previous research, but rather in providing an integrated assessment of their relative importance within the unique economic structure of GCC economies over the period 1980–2023. Rather than addressing an absence of previous research, the study seeks to clarify the relative importance and robustness of competing determinants of emissions in a hydrocarbon-dependent setting characterized by common regional shocks, strong energy dependence, and ongoing economic transformation.

To achieve this objective, the study employs a multi-method panel-data framework combining variable-selection techniques, fixed-effects estimation with Driscoll–Kraay standard errors, and robustness analyses based on alternative estimators. This approach allows the assessment of emissions determinants while accounting for cross-sectional dependence, heterogeneity, and potential endogeneity concerns.

The study addresses the following research questions:

–

How do energy consumption patterns interact with economic growth to shape emission trajectories in rentier, hydrocarbon-dependent economies?

–

What is the net environmental effect of the GCC’s distinctive urbanization model characterized by low-density sprawl coupled with high infrastructure quality?

–

Under what conditions might governance improvements yield measurable emission reductions in fossil fuel-dependent states?

The findings provide updated evidence relevant to policymakers seeking to balance economic diversification, energy transition, and climate commitments across the GCC region.

The paper proceeds as follows: Section 2 reviews the literature on GCC emission drivers, Section 3 details our data and methodology, Section 4 presents empirical results, Section 5 discusses policy implications, and Section 6 concludes with recommendations and future directions.

2. Literature Review

The relationship between economic growth, energy consumption, and carbon emissions has been extensively studied, particularly in fossil fuel-dependent economies such as those of the Gulf Cooperation Council (GCC). The region’s heavy reliance on hydrocarbons—accounting for 60–90% of government revenues (Energy Institute, 2023)—has resulted in some of the world’s highest carbon intensities.

Although numerous studies have examined several of these determinants jointly, the reported findings remain highly heterogeneous across countries, periods, and econometric methodologies. Consequently, important questions remain regarding the relative importance of energy consumption, economic growth, urbanization, trade openness, and institutional quality in shaping emissions dynamics within GCC economies. This section synthesizes the literature around five interrelated channels of emissions: (1) energy consumption, (2) economic growth (EKC), (3) urbanization, (4) trade openness, and (5) institutional quality.

2.1. Energy Consumption and Fossil Fuel Dependence

The relationship between energy consumption and carbon emissions represents one of the most extensively studied topics in environmental economics. A broad consensus exists that fossil-fuel-based energy consumption constitutes a primary source of CO₂ emissions, particularly in resource-rich economies (Apergis & Payne, 2014; Shahbaz et al., 2015). In the GCC region, this relationship is especially pronounced due to the dominance of hydrocarbons in electricity generation, industrial production, and transportation systems.

The GCC economies remain heavily dependent on fossil fuels, with hydrocarbons accounting for a substantial share of both energy production and government revenues. High levels of energy subsidies, energy-intensive industrial activities, and extreme climatic conditions requiring extensive cooling demand contribute to elevated energy consumption and emissions levels. Previous studies consistently report a positive relationship between energy use and environmental degradation in GCC countries (Charfeddine & Kahia, 2019; Alkasasbeh et al., 2023). Recent GCC evidence further confirms that energy intensity and carbon emissions remain closely linked to economic activity and environmental pressures across the region (Abid et al., 2024).

While the positive energy–emissions nexus is well established, the estimated magnitude of this relationship varies considerably across studies and countries. Recent evidence suggests that the environmental impact of energy consumption depends on broader structural factors, including energy-transition policies, renewable-energy deployment, technological progress, and institutional quality. Consequently, evaluating energy consumption alongside other economic and institutional determinants remains important for understanding environmental performance in hydrocarbon-dependent economies.

2.2. Economic Growth and the EKC Hypothesis

The Environmental Kuznets Curve (EKC) hypothesis proposes an inverted U-shaped relationship between economic growth and environmental degradation, whereby pollution initially increases with income before declining after a certain development threshold is reached (Grossman & Krueger, 1995).

Empirical evidence regarding the EKC remains mixed. Some studies support the existence of an inverted U-shaped relationship (Shahbaz et al., 2018), while others report a monotonic positive relationship between income and emissions (Dogan & Turkekul, 2016). The inconsistency of findings is particularly evident in hydrocarbon-dependent economies, where economic expansion often remains closely linked to fossil-fuel extraction and energy-intensive activities.

In the GCC context, studies have produced conflicting results. Mrabet and Alsamara (2017) identify partial support for the EKC hypothesis, whereas other studies conclude that continued hydrocarbon dependence prevents the emergence of a clear turning point (Al-Maamary et al., 2017; Al-Taei et al., 2024).

The absence of a clear consensus in the GCC literature suggests that the income–emissions relationship remains sensitive to model specification, sample period, and country-specific characteristics. Additional evidence is therefore useful for assessing whether economic growth contributes to environmental improvement or environmental degradation within hydrocarbon-dependent economies.

2.3. Urbanization and Built Environment

Urbanization influences environmental quality through multiple channels. Rapid urban expansion may increase emissions through greater energy consumption, transportation demand, and industrial concentration. Conversely, urban density can improve energy efficiency through economies of scale, public transportation systems, and shared infrastructure (Glaeser & Kahn, 2010).

The GCC countries exhibit some of the highest urbanization rates in the world, with most populations concentrated in large metropolitan areas. Previous studies report mixed findings regarding the urbanization–emissions nexus. Some studies find that urbanization increases environmental pressure through infrastructure expansion and transportation activity, while others suggest that compact urban development can contribute to lower emissions intensity (Alawadi & Benkraouda, 2017; Al-Mohannadi et al., 2023).

The absence of consensus indicates that the environmental consequences of urbanization depend on urban planning, infrastructure quality, technological adoption, and demographic characteristics. Consequently, the direction and magnitude of the urbanization–emissions relationship remain largely empirical questions in GCC countries.

2.4. Trade Openness and Industrial Structure

Trade openness affects environmental quality through scale, composition, and technique effects (Grossman & Krueger, 1995). Increased trade can stimulate economic activity and industrial production, leading to higher emissions through the scale effect. Conversely, trade may facilitate access to cleaner technologies and more efficient production processes through the technique effect.

Empirical evidence remains inconclusive. Some studies report that trade openness increases environmental degradation, particularly in countries specializing in pollution-intensive industries (Sbia et al., 2017). Other studies find that international trade promotes environmental improvement through technology transfer and enhanced efficiency (Shahbaz et al., 2015).

Within GCC economies, the environmental impact of trade is particularly complex because exports remain heavily concentrated in hydrocarbon-related sectors. This complexity is further reinforced by the interaction between economic complexity, technological development, and environmental sustainability observed in Saudi Arabia, suggesting that the environmental consequences of economic integration depend on the structure and sophistication of productive activities (Abid & Gafsi, 2025). As a result, existing studies report mixed findings, and no clear consensus has emerged regarding whether trade openness contributes to environmental degradation or environmental improvement in the region.

2.5. Institutional Quality and Climate Governance

Institutional quality is widely recognized as a key determinant of environmental performance. Strong institutions can enhance regulatory enforcement, improve policy implementation, strengthen environmental governance, and facilitate the adoption of sustainable development strategies.

A substantial body of empirical literature finds that governance quality and institutional effectiveness contribute positively to environmental sustainability. Danish et al. (2020) demonstrate that better governance improves environmental quality, while Hassan et al. (2018) show that institutional quality plays a significant role in reducing environmental degradation. Similarly, Baloch et al. (2020) report that stronger institutions contribute to lower ecological footprints and improved environmental outcomes.

However, the literature also reveals considerable heterogeneity in governance effects. The magnitude and significance of institutional variables differ across countries, governance indicators, and environmental measures. In resource-dependent economies, institutional reforms may not immediately translate into environmental improvements because economic incentives remain closely tied to hydrocarbon production and energy-intensive growth models.

In the GCC region, governments have launched several environmental initiatives, including Saudi Vision 2030, the Saudi Green Initiative, and the UAE Net Zero 2050 Strategy. While these initiatives demonstrate increasing institutional commitment to environmental sustainability, empirical evidence regarding their effectiveness remains limited.

Overall, the literature generally supports a role for institutional quality in shaping environmental outcomes. Nevertheless, the magnitude and significance of governance effects vary substantially across indicators, countries, and estimation approaches, suggesting that the environmental effectiveness of institutions remains context-dependent.

2.6. Synthesis and Research Gap

The existing literature provides extensive evidence on the environmental effects of energy consumption, economic growth, urbanization, trade openness, and institutional quality. Numerous studies have examined these variables jointly in both developed and developing economies, including GCC countries. Nevertheless, empirical findings remain highly heterogeneous, with the estimated magnitude, direction, and significance of these determinants varying substantially across countries, periods, and econometric methodologies.

In particular, three unresolved issues remain. First, despite broad agreement regarding the importance of energy consumption, there is no consensus concerning the relative contribution of economic growth, urbanization, trade openness, and institutional quality once these factors are considered simultaneously within hydrocarbon-dependent economies. Second, empirical evidence regarding the Environmental Kuznets Curve (EKC) hypothesis in GCC countries remains inconclusive, with studies reporting positive, negative, U-shaped, and inverted U-shaped relationships. Third, although institutional quality is generally found to improve environmental performance, its effectiveness in resource-dependent economies remains uncertain because environmental policies often coexist with fossil-fuel-based development strategies.

Accordingly, the contribution of this study does not lie in introducing new explanatory variables, as these determinants have been widely examined in previous research. Rather, the study contributes by evaluating their relative importance and robustness within the specific context of GCC economies using a structured multi-method panel-data framework over the period 1980–2023. By comparing the results across complementary estimation techniques, the study seeks to clarify the extent to which alternative determinants consistently explain emissions in hydrocarbon-dependent economies.

3. Methodology

3.1. Data

The study employs an unbalanced panel dataset covering the six Gulf Cooperation Council (GCC) countries—Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, and the United Arab Emirates—over the period 1980–2023. The panel is unbalanced because data availability differs across countries and variables, particularly for the regulatory quality indicator and several early-year observations.

The estimation procedures employed in this study, including fixed-effects estimation with Driscoll–Kraay standard errors, LASSO variable selection, IV-FE estimation, and CCEMG estimation, accommodate unbalanced panels provided that missing observations are not systematically related to the dependent variable. To minimize potential bias, all estimations were performed using the maximum available observations, and the results were compared across alternative specifications to assess robustness.

Although the panel is unbalanced, the proportion of missing observations remains limited relative to the overall sample size and primarily reflects data availability constraints rather than systematic omissions. Consequently, the risk of substantial estimation bias arising from the unbalanced structure is expected to be limited.

The description of variables is presented in Table 1.

Table 1. Data description of GCC Variables.

Variable	Classification	Description	Expected Relationship with CO2 Emissions
CO2E	Environmental	Carbon dioxide (CO2) emissions excluding LULUCF per capita (t CO2E/capita)	Dependent Variable
ENE	Energy-related	Energy use per capita (kg of oil equivalent)	Positive (Higher energy consumption → More emissions)
GDPPC	Economic	GDP per capita (constant 2015 US$)	Non-linear (EKC hypothesis: Inverted U-shape)
DEN	Demographic	Population density (people per sq. km)	Ambiguous (Denser areas may have efficiency gains or pollution congestion)
TRA	Trade-related	Trade as % of GDP	Ambiguous (Trade can increase emissions via production or reduce them via efficiency gains)
REGQ	Institutional	Regulatory Quality (Percentile Rank, 0–100)	Negative (Better regulation → Lower emissions)

The dependent variable is Carbon dioxide (CO₂) emissions per capita (t CO₂E/capita). The independent variables can be categorized into several groups based on their economic and theoretical relevance.

Energy-related factors are represented by energy use per capita (ENE), measured in kilograms of oil equivalent. This variable is expected to exert a positive effect on emissions due to the structural dependence of GCC economies on fossil fuel-intensive energy systems. Economic development is proxied by GDP per capita (GDPPC) measured in constant 2015 US dollars. To formally evaluate the Environmental Kuznets Curve (EKC) hypothesis, both the linear and quadratic forms of GDP per capita are incorporated in alternative model specifications.

Demographic characteristics are represented by population density (DEN), measured as people per square kilometer. Its theoretical effect remains ambiguous, as higher density may either increase emissions through congestion effects or reduce them through infrastructure efficiency and economies of scale.

Trade openness (TRA), measured as trade as a percentage of GDP, is included to capture the environmental effects of international trade integration. Its impact may be either positive or negative depending on the relative dominance of scale, composition, and technique effects. Institutional quality is proxied by the Regulatory Quality index (REGQ), expressed as a percentile rank ranging from 0 to 100.

The inclusion of REGQ is theoretically motivated by environmental governance literature, which emphasizes the role of institutional effectiveness in shaping environmental policy implementation and regulatory enforcement.

All continuous variables are transformed into natural logarithms to reduce heteroskedasticity, improve distributional properties, and facilitate elasticity-based interpretation of estimated coefficients. DDEN and DREGQ are entered in first differences because CIPS indicates they are I(1), whereas the remaining variables are stationary in levels to ensure stationarity and reduce the risk of spurious regression outcomes.

Overall, the selected variables provide a theoretically grounded and empirically tractable framework for analyzing the key structural drivers of CO₂ emissions in GCC economies.

3.2. Econometric Modelling

The empirical strategy is intentionally structured around a limited number of complementary techniques. The fixed-effects model with Driscoll–Kraay standard errors serves as the primary estimation framework. LASSO is employed as an exploratory variable-selection tool to assess the relative relevance of candidate predictors, while IV-FE and CCEMG estimators are used solely as robustness checks to evaluate the stability of the main findings under alternative assumptions regarding endogeneity, heterogeneity, and cross-sectional dependence. This approach prioritizes parsimony and interpretability while maintaining the robustness of the empirical results.

3.2.1. Fixed Effects Regression with Driscoll–Kraay Standard Errors

To control for unobserved country-specific heterogeneity and potential cross-sectional dependence, the study employs a fixed-effects (FE) estimator with Driscoll–Kraay standard errors (Driscoll & Kraay, 1998). This estimator provides standard errors robust to heteroskedasticity, serial correlation, and cross-sectional dependence, making it particularly appropriate for macro-panel settings characterized by strong regional interdependence and common shocks (Hoechle, 2007).

The baseline specification is expressed as:

$$C{O}_2{E}_{it}={\beta }_0+{\beta }_{1}GDPP{C}_{it}+{\beta }_{2}DDE{N}_{it}+{\beta }_{3}TR{A}_{it}+{\beta }_{4}EN{E}_{it}+{\beta }_{5}DREG{Q}_{it}+{\alpha }_i+{\epsilon }_{it}$$

(1)

where ${\alpha }_i$ represents unobserved country-specific effects and ${\epsilon }_{it}$ is the idiosyncratic error term.

The FE–Driscoll–Kraay model is interpreted primarily as capturing contemporaneous associations rather than definitive causal effects.

Variables identified as non-stationary at levels are transformed into first differences where appropriate to ensure stationarity and reduce the risk of spurious regression outcomes. The resulting specifications should be interpreted as capturing contemporaneous associations among the variables rather than long-run equilibrium relationships.

3.2.2. LASSO Regression (Exploratory Variable Selection)

To identify the most relevant predictors and mitigate potential overfitting, we apply the Least Absolute Shrinkage and Selection Operator (LASSO) (Tibshirani, 1996). Unlike conventional regression approaches, LASSO performs simultaneous estimation and variable selection by shrinking less informative coefficients toward zero, thereby improving model parsimony and interpretability.

Given the relatively small cross-sectional dimension of the GCC panel, the LASSO procedure is used primarily as an exploratory variable-selection technique rather than as a structural causal estimator. Although the number of candidate variables is modest, LASSO is used here as a robustness-oriented variable-selection procedure to assess specification uncertainty and relative predictor importance rather than to address high-dimensionality.

The optimal penalty parameter is selected using the Extended Bayesian Information Criterion (EBIC) (Chen & Chen, 2008). This approach ensures a balance between goodness-of-fit and model simplicity, which is particularly important in high-dimensional settings relative to sample size.

Following variable selection, a Post-LASSO Ordinary Least Squares (OLS) regression is estimated to obtain unbiased coefficient estimates and valid statistical inference after removing shrinkage bias introduced during the penalization stage.

Although the number of candidate regressors is limited, LASSO is employed here primarily to evaluate variable relevance, reduce potential specification uncertainty, and complement theory-driven model selection rather than to address high-dimensionality in the conventional machine-learning sense.

The LASSO procedure excluded REGQ from the final selected specification due to its limited explanatory contribution. The revised FE specification is expressed as:

$$C{O}_2{E}_{it}={\beta }_0+{\beta }_{1}EN{E}_{it}+{\beta }_{2}GDPP{C}_{it}+{\beta }_{3}DE{N}_{it}+{\beta }_{4}TR{A}_{it}+{\alpha }_i+{\epsilon }_{it}$$

(2)

To formally examine the EKC hypothesis, an extended specification including the squared GDP term (GDPPC²) is estimated within the same fixed-effects framework.

The quadratic GDP specification is given by:

$$C{O}_2{E}_{it}={\beta }_0+{\beta }_{1}EN{E}_{it}+{\beta }_{2}GDPP{C}_{it}+{\beta }_{3}GDPP{C}_{it}^2+{\beta }_{4}DDE{N}_{it}+{\beta }_{5}TR{A}_{it}+{\alpha }_i+{\epsilon }_{it}$$

(3)

Evidence consistent with the EKC hypothesis is obtained when the coefficient of GDPPC is positive and the coefficient of GDPPC² is negative and statistically significant.

Although only one variable (REGQ) was ultimately excluded, the LASSO procedure serves an important diagnostic role by providing an objective assessment of variable relevance and specification uncertainty. Given the relatively small set of candidate regressors, the LASSO results are reported primarily as supplementary evidence supporting the robustness of the baseline specification rather than as the principal empirical contribution.

3.2.3. Diagnostic and Robustness Tests

Several diagnostic and robustness procedures are conducted to validate the econometric framework and evaluate potential estimation concerns. The Hausman test (Hausman, 1978) is employed to determine the appropriate panel estimator (fixed versus random effects).

Cross-sectional dependence is assessed using the Pesaran (2004) CD test and the Frees test. Given the presence of cross-sectional dependence in several variables, stationarity is evaluated using the Pesaran (2007) Cross-Sectionally Augmented IPS (CIPS) test, which explicitly accounts for cross-sectional dependence through cross-sectional averages. The CIPS procedure is considered more appropriate than first-generation panel unit-root tests because GCC countries are exposed to common economic, financial, and energy-market shocks.

To evaluate multicollinearity, Variance Inflation Factors (VIFs) are computed. Heteroskedasticity is examined using the Breusch–Pagan test, while serial correlation is assessed using the Wooldridge autocorrelation test.

Beyond standard diagnostics, two additional robustness checks are implemented. First, a fixed-effects instrumental-variable (IV-FE) estimator using lagged regressors as instruments is employed to mitigate potential simultaneity and reverse-causality concerns. Second, a Common Correlated Effects Mean Group (CCEMG) estimator is implemented to account for heterogeneous slope coefficients, unobserved common factors, and cross-sectional dependence across GCC countries. Comparing results across these alternative estimators provides additional evidence regarding the stability and robustness of the identified relationships.

Given the relatively small cross-sectional dimension of the GCC panel and the observational nature of the data, the results should be interpreted primarily as robust empirical associations. The use of alternative estimators and diagnostic procedures is intended to strengthen inference and assess the stability of the reported relationships.

4. Results

4.1. Data Analysis

The dataset provides valuable insights into the characteristics of CO₂ emissions and their associated factors across GCC countries.

4.1.1. Descriptive Statistics

The descriptive statistics are summarized and presented in Table 2.

Table 2. Descriptive Statistics.

Variable	Obs.	Mean	Std. Dev.	Min	Max
CO2E	264	3.1707	0.4221	1.975	3.9815
GDPPC	264	10.3338	0.4774	9.426	11.678
DEN	258	4.0173	1.7929	1.0344	7.5663
TRA	226	4.6042	0.3142	3.9063	5.3976
ENE	210	8.9981	0.6109	6.8319	10.004
REGQ	168	4.1603	0.1563	3.6807	4.4304

All variables are log-transformed, with CO₂ emissions (CO₂E) exhibiting moderate variation across the region. GDP per capita (GDPPC) shows relatively stable values, suggesting consistent economic development trajectories among GCC member states. Population density (DEN) displays the widest variation, reflecting substantial heterogeneity between highly urbanized economies and more sparsely populated areas. This heterogeneity in population density is particularly relevant, as it suggests that urbanization effects are unlikely to be uniform across GCC countries and may generate asymmetric impacts on emissions.

Energy consumption (ENE) exhibits consistently high values, confirming the energy-intensive nature of GCC economies. Trade openness (TRA) shows relatively limited variation, indicating persistent integration into global markets. Regulatory quality (REGQ) displays minimal dispersion, suggesting broadly similar institutional frameworks across countries. However, the limited variation in regulatory quality may reduce its explanatory power in econometric models, potentially contributing to its statistical insignificance in subsequent estimations.

These descriptive patterns should be interpreted cautiously given the small cross-sectional dimension and unbalanced observations across variables.

4.1.2. Correlation Analysis

The correlation matrix (Table 3) highlights several important relationships among the variables.

Table 3. Correlation Matrix.

Variable	CO2E	GDPPC	DEN	TRA	ENE	REGQ
CO2E	1.0000
GDPPC	0.7269	1.0000
DEN	0.5335	0.2763	1.0000
TRA	0.2428	0.1419	0.7341	1.0000
ENE	0.8765	0.7509	0.6361	0.3312	1.0000
REGQ	0.2121	0.1991	0.5137	0.6858	0.2536	1.0000

Energy use per capita (ENE) exhibits an exceptionally strong positive correlation with CO₂ emissions (0.8765), confirming its dominant role in driving emissions in oil-dependent economies. GDP per capita (GDPPC) also shows a strong correlation with emissions (0.7269), reinforcing the close linkage between economic growth and energy consumption. The strong correlation between energy consumption and emissions suggests potential concerns regarding multicollinearity and underscores the importance of employing robust estimation techniques.

Regulatory quality (REGQ) displays a positive correlation with emissions (0.2121), which may appear counterintuitive. This pattern likely reflects a structural feature of GCC economies, where improvements in governance have historically coincided with economic expansion rather than environmental regulation, highlighting the limitations of aggregate governance indicators in capturing environmental policy effectiveness. Population density (DEN) and trade openness (TRA) exhibit moderate correlations with emissions and strong interrelationships with each other.

Overall, the correlation structure indicates that emissions in the GCC are shaped by a tightly interconnected system of economic growth, energy use, and urbanization, rather than by isolated factors.

4.1.3. Stationarity Analysis

The CIPS unit root tests (Table 4) indicate a mix of stationary and non-stationary variables.

Table 4. Pesaran CIPS Panel Unit Root Test Results.

Variable	Level CIPS Z(t-bar)	Level p-Value	First Difference CIPS Z(t-bar)	First Difference p-Value
CO2E	−3.016	0.0010	—	—
GDPPC	−1.668	0.0480	—	—
DEN	−0.049	0.4800	−5.313	0.0000
TRA	−2.615	0.0020	—	—
ENE	−2.173	0.0030	—	—
REGQ	0.478	0.6840	−2.945	0.0020

The Pesaran (2007) CIPS test results indicate that CO₂ emissions (CO₂E), GDP per capita (GDPPC), trade openness (TRA), and energy consumption (ENE) are stationary in levels and therefore integrated of order zero, I(0). In contrast, population density (DEN) and regulatory quality (REGQ) are non-stationary in levels but become stationary after first differencing, indicating that they are integrated of order one, I(1). Overall, the results reveal a mixed integration structure consisting of both I(0) and I(1) variables. This finding justifies the use of panel-data estimation techniques that appropriately address cross-sectional dependence and variable-specific stationarity properties. Variables found to be non-stationary at levels are introduced in first differences where appropriate.

The empirical analysis is centered on the fixed-effects model with Driscoll–Kraay standard errors, which constitutes the benchmark specification. The remaining estimators are presented only to assess the robustness of the principal findings rather than as alternative models competing for interpretation.

4.2. Panel Regression Results (FE–Driscoll-Kraay)

The fixed-effects model with Driscoll–Kraay standard errors provides estimates of contemporaneous relationships (Table 5).

Table 5. Regression with Driscoll-Kraay Standard Errors.

Variable	Coefficient	Std. Error	t-Value	p-Value	95% Confidence Interval
GDPPC	0.0879	0.0427	2.0600	0.0550	(−0.0021, 0.1779)
DDEN	−0.0700	0.0134	−5.2100	0.0000	(−0.0983, −0.0416)
TRA	0.1383	0.0215	6.4400	0.0000	(0.0930, 0.1836)
ENE	0.6801	0.0540	12.5900	0.0000	(0.5662, 0.7941)
DREGQ	0.0549	0.0820	0.6700	0.5120	(−0.1181, 0.2280)
_cons	−4.2443	0.3433	−12.3600	0.0000	(−4.9687, −3.5200)
Model Statistics
R-squared	0.8952
F-statistic	191.43 (p = 0.0000)

Energy consumption (ENE) emerges as the most significant driver (0.6801, p < 0.001), indicating that a 1% increase in energy use is associated with a 0.68% increase in emissions. This magnitude confirms the presence of strong energy-emissions elasticity, consistent with the continued dependence of GCC economies on fossil fuel-intensive energy systems.

Trade openness (TRA) shows a significant positive effect (0.1383, p < 0.001), suggesting that increased trade activity is associated with higher emissions. Population density (DDEN) exhibits a negative and significant coefficient (−0.0700, p < 0.001). This result is consistent with the urban efficiency hypothesis, although the relationship should be interpreted as associative and context-dependent.

GDP per capita (GDPPC) is marginally significant (p = 0.055), suggesting a positive association between economic development and emissions. However, this specification does not provide direct evidence regarding the Environmental Kuznets Curve hypothesis. Regulatory quality (DREGQ) is statistically insignificant.

The high explanatory power (R² = 0.8952) suggests that the model captures most within-country variation; however, this should be interpreted cautiously given the small sample size.

4.3. LASSO Analysis

Table 6 shows the results of LASSO variable selection for the regression model where CO₂E (CO₂ emissions) is the dependent variable.

Table 6. Variable Selection via LASSO.

LASSO Variable Selection (EBIC-Optimized Lambda = 0.2360)
Knot	ID	Lambda	s	L1-Norm	EBIC	R2	Action
1	1	82.8725	0	0.0000	−181.5576	0.0000	Added _cons
2	2	75.5103	1	0.0751	−195.2307	0.1615	Added ene
3	11	32.6866	2	0.5129	−341.747	0.8040	Added gdppc
4	33	4.2216	3	0.8622	−532.7504	0.9701	Added tra
5	47	1.1477	4	0.9957	−554.5755	0.9768	Added dden
Selected Variables (λ = 0.2360)
Variable	Lasso Coef.		Post-OLS Coef.		Selection Order
ENE	0.7299		0.7338		1st (strongest)
GDPPC	0.2007		0.2026		2nd
DDEN	−0.2208		−0.2776		3rd
TRA	−0.1112		−0.1178		4th
_cons	−5.0013		−5.0235		-

The LASSO results identify ENE, GDPPC, DDEN, and TRA as the most relevant predictors, while excluding regulatory quality. The exclusion of regulatory quality indicates that this variable contributes relatively little explanatory power within the selected specification. This finding should not be interpreted as evidence that institutions are unimportant for environmental outcomes, but rather that regulatory quality does not emerge as a dominant predictor once the other variables are considered simultaneously. Energy use is selected as the strongest predictor, followed by GDP per capita, confirming their dominant roles.

Importantly, LASSO provides a data-driven validation of the theoretical framework, suggesting that structural economic variables exhibit stronger predictive relevance than institutional indicators within the selected model framework.

Cross-validation results further confirm model robustness.

To validate the model’s generalizability and select a more robust tuning parameter (λ) by testing predictive performance on unseen data, cross-validation is performed after LASSO (Table 7).

Table 7. Cross-Validation Results.

Optimal Lambda (λ = 0.3760)
Variable	Lasso Coef.	Post-OLS Coef.
GDPPC	0.1996	0.2026
DDEN	−0.1872	−0.2776
TRA	−0.1074	−0.1178
ENE	0.7276	0.7338
_cons	−4.9882	−5.0235
Lambda within 1 SE (λ = 2.0060)
Variable	Lasso Coef.	Post-OLS Coef.
GDPPC	0.1911	0.1964
TRA	−0.0625	−0.1173
ENE	0.7037	0.7297
_cons	−4.8981	−4.9377

Cross-validation results indicate that when minimizing prediction error (λ = 0.376), all key variables are retained, with only slight coefficient shrinkage, confirming their robustness and strong predictive power (R² = 0.9768). In contrast, applying the one-standard-error rule yields a more parsimonious model, excluding population density (DDEN) and retaining only GDPPC, TRA, and ENE, suggesting that density adds limited incremental predictive value. Although this simpler model slightly reduces explanatory power, it enhances generalizability and stability.

Overall, the findings highlight energy consumption as the most robust determinant of emissions, while trade and economic growth also play significant roles. From a policy perspective, improving energy efficiency and promoting sustainable transport appear most effective, whereas urbanization effects remain indirect.

The Post-LASSO OLS provides unbiased estimates and proper inference for variables selected by LASSO while maintaining model simplicity (Table 8).

Table 8. Post-LASSO OLS Estimation Results.

Variable	Coefficient	Std. Error	t-Value	p-Value	95% Confidence Interval
GDPPC	0.3341	0.0403	8.28	0.0000	[0.2544, 0.4137]
DDEN	−0.8811	0.2685	−3.28	0.0010	[−1.4111, −0.3510]
TRA	0.1329	0.0445	2.98	0.0030	[0.0449, 0.2208]
ENE	0.4955	0.0283	17.51	0.0000	[0.4396, 0.5514]
_cons	−5.3222	0.3596	−14.8	0.0000	[−6.0323, −4.6122]
Model Fit Statistics
Metric	Value
R-squared	0.8897
Adj-R squared	0.8871
Root MSE	0.1503
AIC	−157.8677
BIC	−142.1594

The Post-LASSO OLS results confirm the statistical significance of all selected variables. Energy use (ENE) remains the strongest determinant (coef = 0.4955), followed by GDP per capita (0.3341). Population density maintains a strong negative effect. The increase in coefficient magnitudes relative to LASSO estimates reflects the removal of shrinkage bias, illustrating the trade-off between bias reduction and variance control.

The consistency between LASSO selection and OLS significance strengthens confidence in the stability and robustness of the identified relationships.

4.4. Diagnostic Tests

To further validate the model, we conduct a formal multicollinearity check using Variance Inflation Factors (VIF) (Table 9).

Table 9. Multicollinearity Diagnostics (VIF Results).

Variable	VIF	1/VIF
ENE	2.35	0.4258
GDPPC	2.14	0.4677
TRA	1.29	0.7775
DEN	1.11	0.8978
Mean VIF	1.72

Diagnostic tests confirm the reliability of the model. The VIF results (mean VIF = 1.72) indicate no multicollinearity concerns, ensuring stable coefficient estimates.

To ensure the robustness and reliability of the regression results, standard diagnostic tests were conducted (Table 10).

Table 10. Diagnostic tests.

Null Hypothesis (H0)	Test Statistic	p-Value
Constant variance (homoskedasticity)	78.0600	0.0000
No first-order autocorrelation	1.8510	0.2317

The Wooldridge test shows no autocorrelation (p = 0.2317), while the Breusch–Pagan test reveals heteroskedasticity, justifying the use of robust estimators.

Although the Wooldridge test does not detect first-order autocorrelation, this result should be interpreted cautiously given the relatively small cross-sectional dimension of the panel.

The examination of cross-sectional dependence (Pesaran CD and Frees tests) in the GCC panel data is shown in Table 11.

Table 11. Cross-Sectional Dependence Test Results.

Pesaran CD Test
Variable	Test Statistic	p-Value
CO2E	−0.4840	0.6280
GDPPC	−0.5230	0.6010
DEN	5.7060	0.0000
TRA	7.7440	0.0000
ENE	6.7830	0.0000
Frees Test of Cross-Sectional Independence
Test Statistic	Critical Values (α)
0.7400	0.184 (10%)
	0.243 (5%)
	0.360 (1%)

Cross-sectional dependence tests indicate no dependence for CO₂ emissions and GDP per capita, but strong cross-sectional interdependence for DEN, TRA, and ENE, reflecting common regional shocks, economic integration, and shared energy-market dynamics across GCC countries.

These findings further justify the reliance on heteroskedasticity- and dependence-robust estimators throughout the empirical analysis.

4.5. Revised FE Model (Robust Errors)

To address cross-sectional dependence and potential heteroskedasticity, we re-estimate the fixed-effects model using Driscoll–Kraay standard errors (Table 12).

Table 12. Fixed-Effects Regression with Driscoll-Kraay Standard Errors.

Variable	Coefficient	Std. Error	t-Value	p-Value	95% Confidence Interval
GDPPC	0.2366	0.0692	3.4200	0.0020	[0.0958; 0.3773]
DDEN	−0.0590	0.0224	−2.6400	0.0130	[−0.1045; −0.0135]
TRA	0.4155	0.0771	5.3900	0.0000	[0.2587; 0.5722]
ENE	0.3506	0.0912	3.8500	0.0010	[0.1653; 0.5358]
_cons	−4.1313	0.4251	−9.7200	0.0000	[−4.9953; −3.2673]
Model Fit Statistics
Metric	Value
R-squared	0.5915
F-test	92.46 (p = 0.000)

Accordingly, subsequent revised fixed-effects specifications prioritize the subset of variables selected through the LASSO procedure, while alternative robustness specifications retain theoretically relevant variables where conceptually appropriate. This approach allows us to combine data-driven variable selection with robust panel estimation, improving both model parsimony and inferential reliability.

The revised model confirms the robustness of the main findings. GDP per capita and energy use remain positive and significant, while population density remains negative but less significant. The reduction in statistical significance for some variables reflects the correction for cross-sectional dependence, suggesting that earlier estimates may have been upwardly biased.

4.6. Fixed-Effects Regression with GDP-Squared Term: Formal EKC Specification

To further examine the Environmental Kuznets Curve (EKC) hypothesis, the model was re-estimated by including a quadratic GDP per capita term (GDPPC²) within the fixed-effects framework using Driscoll–Kraay standard errors (Table 13).

Table 13. EKC Robustness Test: Fixed-Effects Regression with Driscoll–Kraay Standard Errors.

Variable	Coefficient	Std. Error	t-Statistic	p-Value	95% Confidence Interval
GDPPC	−4.5636	2.4429	−1.8700	0.0700	[−9.5282; 0.4009]
GDPPC2	0.2309	0.1189	1.9400	0.0610	[−0.0108; 0.4726]
DDEN	−0.0805	0.0249	−3.2300	0.0030	[−0.1312; −0.0298]
TRA	0.3770	0.0738	5.1100	0.0000	[0.2270; 0.5270]
ENE	0.4150	0.0737	5.6300	0.0000	[0.2652; 0.5648]
_cons	20.4523	12.6864	1.6100	0.1160	[−5.3296; 46.2342]
Model Fit Statistics
Metric	Value
within R-squared	0.6154
F-test	94.60 (p = 0.000)
Turning Point Analysis	Value	Std. Error	z-Statistic	p-Value	95% Confidence Interval
Estimated Turning Point (GDP per capita)	19,561.38 USD	4635.9100	4.2200	0.0000	[10,475.17; 28,647.59]

Notes: The dependent variable is log CO₂ emissions per capita. GDPPC² is included to test the non-linear Environmental Kuznets Curve specification. Driscoll–Kraay standard errors are robust to heteroskedasticity, autocorrelation, and cross-sectional dependence. The coefficient of GDPPC is negative while the coefficient of GDPPC² is positive, suggesting a U-shaped rather than an inverted U-shaped relationship. However, both GDPPC and GDPPC² are only marginally significant at the 10% level and not significant at the 5% level. Therefore, the results do not provide strong evidence in favor of the conventional EKC hypothesis.

The results reveal that the coefficient of GDP per capita is negative and marginally significant at the 10% level (−4.5636, p = 0.070), while the coefficient of the squared GDP term is positive and also marginally significant at the 10% level (0.2309, p = 0.061).

These findings suggest a possible U-shaped income–emissions relationship. However, because both GDP coefficients are only marginally significant at the 10% level and fail to achieve significance at the 5% level, the evidence should be regarded as suggestive rather than conclusive. In other words, emissions initially decline at lower income levels but begin to increase again as income rises. However, because both coefficients are only weakly significant and not statistically significant at the 5% level, the evidence remains insufficient to confirm a robust non-linear income–emissions relationship.

The estimated turning point is approximately 19,561 USD GDP per capita. Nevertheless, this threshold should be interpreted cautiously because the estimated coefficients do not support the conventional EKC pattern. Under the traditional EKC hypothesis, GDP should exhibit a positive coefficient while the squared GDP term should be negative, indicating that emissions eventually decline after reaching a certain income level. In contrast, the present findings indicate the opposite pattern, suggesting that emissions may rise again beyond the estimated income threshold. Therefore, the calculated turning point does not represent evidence of a transition toward lower emissions at higher stages of economic development.

Overall, the results do not provide robust empirical support for the conventional inverted-U Environmental Kuznets Curve hypothesis in GCC countries. Instead, they suggest that economic growth in fossil fuel-intensive economies may continue to generate environmental pressures even at relatively advanced income levels, reflecting the persistent dependence of GCC countries on fossil fuel-intensive production structures and energy consumption patterns.

The positive and highly significant coefficients of trade openness and energy consumption further reinforce this interpretation. Energy use remains one of the strongest determinants of emissions, indicating that economic expansion in GCC countries is still closely linked to carbon-intensive energy systems. Meanwhile, population density exhibits a negative and significant coefficient, suggesting that urban concentration may generate efficiency gains capable of partially mitigating emissions growth.

4.7. Cointegration Analysis

Given the presence of cross-sectional dependence, the Westerlund (2007) error-correction-based panel cointegration test was employed as a second-generation cointegration procedure. Unlike first-generation tests, the Westerlund framework is more appropriate in panels characterized by cross-sectional dependence because it is based on the significance of the error-correction mechanism rather than residual stationarity.

The results are reported in Table 14.

Table 14. Westerlund (2007) Cointegration Test Results.

Statistic	Value	Z-Value	p-Value
Gt	−1.691	1.222	0.889
Ga	−3.032	2.773	0.997
Pt	−2.856	1.434	0.924
Pa	−2.148	1.812	0.965

These findings suggest that there is no evidence of a stable long-run equilibrium relationship among CO₂ emissions, economic growth, population density, trade openness, energy consumption, and regulatory quality in the GCC countries over the study period. Consequently, the results do not support the existence of panel cointegration among the variables.

Given the absence of cointegration, subsequent analyses focus primarily on contemporaneous and medium-run relationships estimated through fixed-effects models with Driscoll–Kraay standard errors and other robustness estimators. Accordingly, any long-run interpretations should be treated with caution.

4.8. Robustness Checks

4.8.1. Endogeneity Robustness: Fixed-Effects IV Regression

An instrumental-variable fixed-effects estimator was additionally employed to mitigate potential endogeneity concerns associated with energy consumption, GDP per capita, and trade openness. Lagged values of the potentially endogenous regressors were used as instruments (Table 15).

Table 15. Fixed-Effects Instrumental Variable (IV-FE) Regression.

Variable	Coefficient	Robust Std. Error	z-Statistic	p-Value	95% Confidence Interval
ENE	0.4733	0.0616	7.6800	0.0000	[0.3526; 0.5941]
GDPPC	0.1596	0.0444	3.5900	0.0000	[0.0725; 0.2467]
TRA	0.3601	0.2268	1.5900	0.1120	[−0.0843; 0.8046]
DEN	−0.0831	0.0357	−2.3300	0.0200	[−0.1531; −0.0131]
Constant	−4.0950	0.4816	−8.5000	0.0000	[−5.0388; −3.1511]
Model Statistics
Statistic	Value
Within R2	0.6078
Between R2	0.8831
Overall R2	0.8117
Wald χ2	1186.4200
Prob > χ2	0.0000
Rho	0.7802
Instrumented variables: ENE, GDPPC, TRA
Instruments: DEN, L.ENE, L.GDPPC, L.TRA

Notes: The IV fixed-effects estimator was implemented to mitigate potential endogeneity arising from simultaneity and reverse causality between emissions, economic growth, trade openness, and energy consumption.

Prior to interpreting the IV-FE estimates, several diagnostic tests were conducted to assess instrument relevance, identification, and validity (Table 16). The Kleibergen–Paap rk LM statistic (22.110, p < 0.001) rejects the null hypothesis of under-identification, confirming that the instruments are sufficiently correlated with the endogenous regressors. The Kleibergen–Paap rk Wald F statistic (15.541), together with the first-stage Sanderson–Windmeijer statistics, indicates that the excluded instruments are not weak.

Table 16. IV-FE Diagnostic Tests.

Test	Statistic	p-Value	Conclusion
Kleibergen–Paap rk LM (Under-identification)	22.1100	0.0000	Reject H0: Model is identified
Kleibergen–Paap rk Wald F (Weak identification)	15.5410	—	Instruments are sufficiently strong
Hansen J (Over-identification)	0.0000	0.9918	Fail to reject H0: Instruments are valid
Durbin–Wu–Hausman Endogeneity Test	11.8070	0.0081	Reject H0: Endogeneity exists

Because the model is overidentified, with three endogenous regressors and four excluded instruments, the Hansen J test was employed to evaluate instrument validity. The Hansen J statistic (p = 0.9918) fails to reject the null hypothesis that the instruments are exogenous, supporting the validity of the instrument set. Finally, the Durbin–Wu–Hausman endogeneity test rejects the null hypothesis of exogeneity (χ² = 11.807, p = 0.0081), confirming that potential endogeneity cannot be ruled out and supporting the use of instrumental-variable estimation as a robustness exercise.

Overall, the diagnostic evidence indicates that the IV-FE specification is properly identified, free from weak-instrument concerns, and based on a valid set of instruments.

The IV estimates remain broadly consistent with the baseline findings, particularly regarding the positive impact of energy consumption and economic growth on CO₂ emissions and the negative effect of population density. This consistency suggests that the main conclusions are robust to potential simultaneity bias and reverse causality.

4.8.2. Cross-Sectional Dependence Robustness: CCEMG/CCE Mean Group

To further account for cross-sectional dependence and heterogeneous slope coefficients across GCC countries, a Common Correlated Effects Mean Group (CCEMG) estimator was implemented as a second-generation panel robustness check (Table 17).

Table 17. Common Correlated Effects Mean Group (CCEMG) Estimation.

Variable	Coefficient	Std. Error	z-Statistic	p-Value	95% Confidence Interval
GDPPC	0.0875	0.1870	0.4700	0.6400	[−0.2790; 0.4540]
DEN	−0.2317	0.3101	−0.7500	0.4550	[−0.8394; 0.3761]
TRA	0.0826	0.0153	5.4000	0.0000	[0.0526; 0.1126]
ENE	0.5509	0.1094	5.0400	0.0000	[0.3366; 0.7653]
Cross-sectional average of CO2E	0.7904	0.2932	2.7000	0.0070	[0.2157; 1.3651]
Cross-sectional average of GDPPC	0.0606	0.0442	1.3700	0.1700	[−0.0260; 0.1472]
Cross-sectional average of DEN	0.3221	0.3729	0.8600	0.3880	[−0.4088; 1.0530]
Cross-sectional average of TRA	−0.1033	0.1679	−0.6100	0.5390	[−0.4324; 0.2258]
Cross-sectional average of ENE	−0.6737	0.2975	−2.2600	0.0240	[−1.2568; −0.0905]
Constant	−0.2501	2.6546	−0.0900	0.9250	[−5.4530; 4.9527]
Model Statistics
Statistic	Value
Wald χ2	55.2800
Prob > χ2	0.0000
RMSE	0.0583

Notes: The CCEMG estimator accounts for cross-sectional dependence, heterogeneous slope coefficients, and unobserved common factors across GCC countries. Cross-sectional averages are included to capture regional common shocks and interdependence effects.

The CCEMG estimates continue to identify energy consumption and trade openness as significant determinants of emissions after accounting for cross-sectional dependence and unobserved common factors. The loss of significance for GDP per capita and population density suggests that these relationships may be more sensitive to model specification and regional common shocks.

Taken together, the FE–Driscoll–Kraay, IV-FE, and CCEMG estimators produce broadly consistent evidence regarding the dominant role of energy consumption in explaining CO₂ emissions in GCC countries. While the significance of certain variables varies across estimators, the positive influence of energy use remains remarkably stable, reinforcing the conclusion that fossil fuel dependence continues to be the primary driver of environmental pressure in the region.

The predominance of energy consumption over institutional factors is broadly consistent with recent evidence from GCC and other emerging economies, which highlights the persistence of energy-driven emissions dynamics even during periods of economic diversification and environmental policy reform (Abid, 2026).

5. Discussion

The findings should be interpreted in light of the distinctive characteristics of GCC economies, which combine high hydrocarbon dependence, energy-intensive production structures, extensive energy subsidies, and ongoing economic diversification programs. Across all estimators, energy consumption emerges as the dominant determinant of emissions, suggesting that improvements in energy efficiency and the expansion of renewable energy remain central to achieving decarbonization objectives. In contrast, institutional quality exhibits limited explanatory power once structural economic variables are controlled for, indicating that governance improvements alone may be insufficient to reduce emissions unless accompanied by structural changes in energy production and consumption patterns.

The findings of this study provide important insights into the determinants of CO₂ emissions in GCC countries over the period 1980–2023. The analysis employs a structured multi-method framework combining fixed-effects estimation with Driscoll–Kraay standard errors, LASSO variable selection, and robustness analyses based on IV-FE and CCEMG estimators. Each method serves a complementary purpose, allowing the assessment of emissions determinants while accounting for cross-sectional dependence, heterogeneity, and potential endogeneity concerns. The results provide updated evidence on the relative importance of economic, demographic, institutional, and energy-related factors in hydrocarbon-dependent economies.

Energy consumption emerges as the most dominant and consistent driver of CO₂ emissions across virtually all specifications. The baseline fixed-effects regression with Driscoll–Kraay standard errors indicates that a 1% increase in energy consumption is associated with approximately a 0.68% increase in emissions (p < 0.001). The IV-FE and CCEMG estimators yield qualitatively similar results, confirming the robustness of the positive relationship between energy use and emissions. The consistency of these findings across alternative specifications suggests that GCC economies remain structurally dependent on fossil fuel-intensive energy systems (Al-Mulali & Ozturk, 2015; Saudi et al., 2018).

The present findings further support the “carbon lock-in” argument developed in the energy-transition literature, whereby historical investment in hydrocarbon infrastructure creates persistent dependence on fossil fuel-intensive production systems (Unruh, 2000). This dependence likely reflects a combination of fossil fuel subsidies, energy-intensive industrialization patterns, and climatic conditions requiring substantial cooling demand throughout the year (Krane, 2018). Consequently, the findings suggest that decarbonization strategies in GCC economies must prioritize energy-system transformation through renewable energy deployment, energy-efficiency improvements, and gradual subsidy reform. In particular, investments in solar, wind, and green hydrogen infrastructure may play a critical role in reducing the region’s carbon intensity and supporting ongoing decarbonization objectives (IEA, 2023).

The relationship between economic growth and CO₂ emissions provides only limited evidence regarding the Environmental Kuznets Curve (EKC) hypothesis. In the baseline fixed-effects specification, GDP per capita exhibits a positive and marginally significant coefficient, suggesting that economic expansion remains associated with rising environmental pressures in GCC countries.

When the quadratic GDP specification is introduced, the coefficient of GDP per capita becomes negative while the coefficient of the squared GDP term becomes positive. These coefficient signs suggest a possible U-shaped relationship rather than the conventional inverted-U EKC pattern. However, both coefficients are only marginally significant at the 10% level and fail to achieve significance at the 5% level. Consequently, the evidence should be interpreted as suggestive rather than conclusive.

The dominant role of energy consumption observed in this study is consistent with recent GCC evidence reported by Chaabouni and Abid (2025), who identify economic activity and structural characteristics as major drivers of energy demand across GCC countries.

Importantly, the FE–Driscoll–Kraay EKC results are only marginally significant at the 10% level, indicating that the evidence remains sensitive to specification and should therefore be interpreted cautiously. Although the FMOLS specification provides stronger statistical significance, the estimated turning points should not be interpreted as confirmation of the conventional EKC hypothesis because the coefficient signs do not support the traditional inverted U-shaped pattern.

These findings are consistent with studies suggesting that the EKC framework may not fully capture environmental dynamics in resource-dependent economies (Ahmad, 2024; Charfeddine & Kahia, 2019; Nasir et al., 2019). This interpretation is also consistent with Abid (2025), who argues that sustainable development outcomes depend on balancing economic growth with long-term environmental and resource sustainability objectives rather than relying on growth alone to improve environmental performance. Taken together, the results suggest that economic growth alone is unlikely to generate automatic environmental improvements in fossil fuel-dependent economies.

Instead, continued economic expansion may reinforce carbon-intensive production structures unless accompanied by substantial structural transformation, technological upgrading, and energy diversification policies.

Trade openness exhibits more ambiguous and model-dependent effects across specifications. The FE–Driscoll–Kraay estimations indicate a positive and statistically significant relationship between trade openness and emissions, suggesting that trade expansion may initially increase emissions through higher production activity, transportation intensity, and energy demand. The revised FE specification and the CCEMG estimator continue to indicate a positive association between trade openness and emissions, although the magnitude of the effect varies across specifications.

However, the significance and magnitude of the trade openness coefficient vary across estimators, indicating that the environmental effects of trade integration are not fully stable across model specifications. Several factors may explain this sensitivity. First, trade openness may simultaneously generate opposing environmental effects through scale, composition, and technique channels. While greater trade activity can increase emissions through higher production, transportation, and energy demand, it may also facilitate access to cleaner technologies, more efficient production processes, and environmentally friendly capital goods. Second, GCC economies exhibit a highly specialized export structure dominated by hydrocarbons, which may amplify the environmental consequences of trade relative to more diversified economies. Third, differences across estimators in accounting for cross-sectional dependence, unobserved common factors, and country-specific heterogeneity may affect the estimated magnitude of the trade effect.

The observed coefficient instability therefore suggests that trade openness should not be interpreted as a uniformly beneficial or harmful determinant of environmental quality in GCC countries. Rather, its environmental impact appears conditional on the structure of trade flows, the technological content of imports and exports, and the broader policy environment within which international integration occurs. Consequently, the trade-related results should be interpreted with greater caution than those associated with energy consumption, which remain highly stable across all specifications.

These findings are consistent with the theoretical framework proposed by Grossman and Krueger (1995), who argue that trade may simultaneously generate scale, composition, and technique effects with opposing environmental consequences. While hydrocarbon exports may intensify emissions through scale effects, trade may also facilitate access to cleaner technologies and energy-efficient capital goods (Shahbaz et al., 2015).

Accordingly, the results neither fully support the “pollution haven” hypothesis nor provide definitive confirmation of the “pollution halo” hypothesis (Frankel & Rose, 2005). Instead, they indicate that trade openness may generate both positive and negative environmental effects depending on institutional capacity, industrial structure, and technological adaptation. From a policy perspective, this highlights the importance of integrating environmental standards and clean-technology provisions into regional trade and industrial strategies.

Population density consistently exhibits a negative relationship with emissions across several specifications, including the FE–Driscoll–Kraay, Post-LASSO OLS, and IV-FE specifications. This pattern is broadly consistent with the urban efficiency hypothesis, according to which higher urban density may be associated with lower per capita emissions through shared infrastructure, reduced transportation intensity, and economies of scale in energy utilization. However, this interpretation should be treated cautiously because the estimated relationship is not uniformly stable across all specifications and does not necessarily imply a causal effect of population density on environmental outcomes (Glaeser & Kahn, 2010). The negative coefficient remains statistically significant in several specifications, suggesting that denser urban structures may be associated with greater environmental efficiency in GCC economies. Nevertheless, the observed relationship is likely influenced by broader contextual factors, including infrastructure quality, transportation systems, urban planning practices, and economic composition.

Nevertheless, the significance of population density weakens in some alternative specifications such as the CCEMG estimator and EKC model, suggesting that the urbanization-emissions relationship may not be fully stable across all econometric frameworks. This variability is consistent with recent urban sustainability studies emphasizing that the environmental effects of urban concentration depend critically on urban planning quality, transport systems, climatic conditions, and infrastructure efficiency rather than density itself (Al-Mohannadi et al., 2023; UN-Habitat, 2020). Consequently, urbanization should not automatically be interpreted as environmentally beneficial. Instead, the results suggest that carefully planned urban development, public transportation systems, and smart-city investments may help mitigate emissions growth in rapidly expanding GCC metropolitan areas.

Regulatory quality does not exhibit a statistically robust relationship with emissions across the empirical specifications. The variable becomes statistically insignificant in the FE–Driscoll–Kraay model and is excluded entirely from the LASSO-selected specification due to its limited predictive contribution. This finding is broadly consistent with previous GCC-focused studies reporting weak or statistically insignificant governance effects in environmental models (Alkasasbeh et al., 2023; Halkos & Tzeremes, 2013). The results suggest that aggregate governance indicators may not adequately capture the effectiveness of environmental regulation and policy enforcement in GCC economies.

Several explanations may account for this result. First, the relatively limited variation in regulatory quality across GCC countries may reduce its explanatory power within panel estimations. Second, formal institutional indicators may fail to reflect actual enforcement capacity or the implementation effectiveness of environmental policies. Third, political economy constraints associated with hydrocarbon dependence may weaken the practical effectiveness of environmental governance despite improvements in formal institutional quality (Stepanov & Makarov, 2024).

The robustness analysis further strengthens confidence in the main findings. The IV-FE estimator indicates that the positive effects of energy consumption and economic growth remain broadly stable after addressing potential simultaneity concerns. Similarly, the CCEMG estimator confirms that energy consumption and trade openness remain significant even after controlling for unobserved common factors and heterogeneous cross-sectional dynamics. These results suggest that the principal conclusions are not driven by a particular estimation approach.

Nevertheless, the robustness results should be interpreted cautiously given the relatively small cross-sectional dimension of the GCC panel. Although alternative estimators produce broadly consistent results, the findings should primarily be viewed as robust empirical associations rather than definitive causal relationships.

Overall, the findings demonstrate that CO₂ emissions in GCC economies are shaped by the interaction of energy use, economic activity, trade integration, urbanization patterns, and institutional factors. The results consistently highlight the dominant role of fossil fuel-intensive energy systems while simultaneously suggesting that urban efficiency gains and technological modernization may partially mitigate emissions growth under appropriate policy conditions. Consequently, achieving sustained decarbonization in GCC economies will likely require coordinated structural reforms targeting energy systems, industrial diversification, urban planning, and environmental governance simultaneously rather than relying on isolated policy interventions.

More broadly, all reported relationships should be interpreted as robust empirical associations rather than definitive causal effects, as the environmental consequences of economic, demographic, and institutional factors remain highly context-dependent within hydrocarbon-dependent economies.

6. Conclusions and Recommendations

This study examines the determinants of CO₂ emissions in Gulf Cooperation Council (GCC) countries over the period 1980–2023, focusing on the roles of energy consumption, economic growth, trade openness, population density, and regulatory quality. Using a structured multi-method panel-data framework combining fixed-effects estimation with Driscoll–Kraay standard errors, LASSO variable selection, and robustness analyses based on IV-FE and CCEMG estimators, the study provides a comprehensive assessment of the factors shaping environmental outcomes in hydrocarbon-dependent economies while accounting for cross-sectional dependence, heterogeneity, and potential endogeneity concerns.

The empirical findings consistently identify energy consumption as the most dominant and robust determinant of CO₂ emissions across nearly all specifications. This result confirms the persistent structural dependence of GCC economies on fossil fuel-intensive energy systems and suggests that energy use remains the principal mechanism through which economic activity translates into environmental pressure. The stability of this relationship across alternative estimators indicates that meaningful reductions in emissions will require substantial transformation of regional energy systems rather than incremental policy adjustments alone.

Economic growth generally exhibits a positive association with emissions, although the EKC specifications provide only limited evidence of a non-linear income–emissions relationship. The estimated coefficients do not support the conventional inverted-U Environmental Kuznets Curve hypothesis. Instead, the results suggest that continued economic expansion may reinforce environmental pressures unless accompanied by structural transformation, economic diversification, and technological upgrading. These findings highlight the limitations of relying on income growth alone to generate environmental improvements in hydrocarbon-dependent economies.

The environmental effects of trade openness appear more conditional and model-dependent. While several specifications indicate that trade expansion contributes positively to emissions through increased production activity, transportation intensity, and energy demand, the magnitude and significance of the estimated effect vary across estimators. This pattern suggests that the environmental consequences of trade integration depend on the balance between scale effects, industrial composition, and technology-transfer mechanisms. Similarly, population density generally exhibits a negative relationship with emissions, providing tentative support for the urban efficiency hypothesis. However, the significance of this relationship weakens in some alternative specifications, indicating that the environmental benefits of urbanization depend on broader infrastructural and planning conditions rather than density itself.

In contrast, regulatory quality does not exhibit a statistically robust relationship with emissions and is excluded from the preferred LASSO specification. This finding suggests that aggregate governance indicators may not adequately capture the effectiveness of environmental regulation and enforcement mechanisms in GCC economies. More broadly, the results imply that governance improvements alone may be insufficient to reduce emissions unless accompanied by structural changes in energy production and consumption patterns.

This study makes three principal contributions to the literature on environmental sustainability in GCC countries. First, from an empirical perspective, it provides updated evidence showing that energy consumption remains the most consistent and influential determinant of CO₂ emissions, whereas the effects of trade openness, population density, and institutional quality are more sensitive to model specification and contextual conditions. Second, from a methodological perspective, the study combines variable-selection techniques with complementary panel-data estimators to assess the relative importance and robustness of competing emissions determinants within a small-sample, hydrocarbon-dependent setting. Third, from a policy perspective, the findings highlight the central role of energy-system transformation while suggesting that economic diversification, sustainable urban development, and stronger environmental governance can complement decarbonization efforts.

From a policy standpoint, the results strongly support the need for coordinated structural reforms aimed at accelerating the energy-transition process across GCC countries. Large-scale investments in renewable energy infrastructure, including solar, wind, and green hydrogen technologies, should constitute a central pillar of regional decarbonization strategies. These efforts should be complemented by gradual fossil fuel subsidy reform, improved energy-pricing mechanisms, and stronger incentives for energy efficiency across industrial, transportation, and construction sectors.

Economic diversification remains equally important for reducing environmental vulnerability associated with hydrocarbon dependence. Expanding knowledge-based industries, green finance initiatives, and low-carbon industrial activities may help decouple economic growth from emissions while improving economic resilience. Regional trade and industrial policies should also encourage clean technology diffusion, sustainable production systems, and environmentally oriented investment frameworks. Urban planning strategies should prioritize compact and transit-oriented development supported by investments in public transportation, energy-efficient buildings, and smart infrastructure systems.

At the institutional level, environmental governance frameworks require further strengthening. The absence of a statistically robust governance effect suggests that environmental policy effectiveness depends not only on formal institutional quality but also on implementation capacity, enforcement mechanisms, monitoring systems, and political commitment to structural reform. Consequently, future policy efforts should focus on improving the operational effectiveness of environmental regulation rather than relying exclusively on aggregate governance indicators.

Several limitations should be acknowledged. First, the analysis focuses exclusively on GCC countries, which may limit the generalizability of the findings to other regions. Second, the panel is unbalanced due to data availability constraints, particularly for regulatory quality and some early-year observations. Third, although multiple estimation techniques are employed, the results should primarily be interpreted as robust empirical associations rather than definitive causal relationships given the observational nature of the data and the relatively small cross-sectional dimension of the sample. Future research may extend the analysis by incorporating renewable-energy indicators, sectoral emissions measures, climate-policy variables, and alternative institutional dimensions. Comparative analyses involving other resource-dependent economies may also provide valuable insights into alternative transition pathways and transferable policy lessons.

Overall, the findings underscore the importance of integrated policy responses simultaneously targeting energy systems, industrial structures, trade patterns, urban development, and environmental governance. Achieving long-term sustainability in GCC economies will require a comprehensive transition away from carbon-intensive growth models toward more diversified, energy-efficient, and environmentally sustainable development pathways.