Decoupling Economic Expansion from Environmental Degradation: A Panel ARDL Analysis of Renewable Energy’s Role in Arab MENA Countries

A substantial corpus of previous research has studied the interactions tying environmental degradation and EG but has ignored the potential influence of green energy use on these interactions. Therefore, such neglect may cause misleading results for some countries in the MENA region, mainly where a series of economic and environmental reforms can induce some permanent structural changes in their sustainable development in the long run. Before developing our predetermined expectations regarding these variables into a cohesive framework, we scrutinize the bilateral interconnection between environmental pollution, RE use, and economic expansion separately.

2.1 The economic growth-environmental pollution nexus

The research investigations exploring the interaction between pollutant emissions (CO₂) and EG in the MENA area remain a contention between researchers. On the one hand, numerous empirical research studies have suggested that EG often enhances the mitigation of CO₂ emissions, which are key indicators of environmental pollution in the MENA region. Among these studies, Ben Cheikh and Ben Zaied [20] carried out one noteworthy research using a nonlinear panel threshold regression framework to reexamine the dynamic interplay between CO₂ pollutant emissions and per capita income in the MENA region. Their results reveal a threshold effect in carbon dioxide evolution, indicating that the influence of income growth follows a nonlinear pattern and is influenced by various energy-related factors. Similarly, to explore the dynamic interactions between economic expansion, energy transition, and technological level in 21 MENA countries from 1997 to 2021, Alariqi et al. [21] applied a Vector Error Correction model (VEC)for the short-run analysis and the ARDL model for the long-run analysis. The authors indicated that as MENA countries experience EG, they tend to direct resources toward environmentally friendly technologies and sustainable practices, which helps minimize pollutant (CO₂) emissions. Omri et al. [22] explored the main factors influencing the CO₂ output for 16 countries in the MENA region between 1990 and 2018. Their study emphasized the capacity of economic development and energy use to influence the dynamics of CO₂ emissions. For this objective, the authors decomposed the economy into a triad of core components: industry, services, and agriculture. The aggregate traditional energy demand was divided into fossil and green energy sources. The authors applied an econometric tool represented by the Augmented Mena Group (AMG) and the Common Correlated Effect Mean Group (CCE-MG). Their empirical findings suggest that a structural shift towards a RE-based economy, driven by EG, can significantly reduce CO₂ emissions.

However, some empirical investigations have indicated a direct interdependence between economic expansion and CO₂ emissions; environmental quality continues to degrade with rising per capita income. in the MENA area. In this context, we can reference the research conducted by Alshehry and Belloumi [23] for 17 MENA countries to examine the asymmetric and symmetric effects of traditional energy use and economic expansion on aggregate CO₂ emissions from 1990 to 2020. To this end, the authors applied a panel linear and nonlinear ARDL estimation approach. They concluded that expanding economic activities in the MENA nations may result in environmental degradation through increased CO₂ discharge in the long term. Meanwhile, increasing energy demand could increase CO₂ emissions in the short-run and long-run.

Additionally, Omri and Saidi [7] assessed the impacts of fossil fuel and green energy sources on CO₂ flows across 14 MENA countries between 1990 and 2019, using vector error correction and fully modified least-squares techniques. Their analysis revealed that fossil energy usage significantly increases CO₂ discharge, while green energy sources have a mitigating effect. However, the overall effect of EG remained detrimental to environmental quality.

To examine the potential impact of excessive usage of traditional energy, urban growth, and economic development on the increase of CO₂ output in Belt and Road countries from 1985 to 2017, Liu et al. [24] applied some econometric techniques, including a panel cointegration approach, Granger causality tests, and heterogeneity analysis. According to their findings, the main factors that negatively affect CO₂ emissions are the increasing use of traditional energy, the expansion of the economy, and the spread of urbanization. Furthermore, according to the panel causality study, their energy usage, GDP growth, urbanization, and CO₂ emissions are all positively correlated.

In the same context, Fodha and Zaghdoud [25] examined the pollutant emissions of Tunisia’s economy from 1961 to 2004 with the EG nexus. Employing Johansen's cointegration analysis, the study is conducted regarding the environmental Kuznets curve theory, with CO₂ emission and GDP serving as the environmental and economic indicators, respectively. The findings demonstrated that, over time, the connection between GDP and CO₂ emissions is monotonically growing. In addition, the causality test findings showed that the link between pollution and income in the Tunisian economy is one-way, with income changing the environment and not the other way around.

Several factors explain the apparent contradictions in the literature. The different econometric estimations (such as panel data, time series analysis, or econometric methods such as ARDL and EKC models) often yield different results. Moreover, sample selection variations greatly influence findings; studies of developed nations often find a decoupling effect between EG and emissions, while studies of developing countries frequently report a positive correlation. Also, temporal differences are essential, with the effects of green technologies and policy reforms more likely to be captured in recent studies than in earlier studies. Understanding how economic development and environmental degradation are related across different settings requires the analysis of these patterns.

2.2 The renewable energy consumption-environmental pollution nexus

The literature reveals a complex and multifaceted interrelation between sustainable energy use and environmental pollution in the MENA area. While some studies indicate that green energy consumption can indeed contribute to diminishing CO₂ output pollution, others suggest that, under specific conditions, its adoption may not always translate into environmental improvement. For this instance, Adun et al. [26] ran a scenario study in Algeria, Morocco, Tunisia, Egypt, and Libya using the Energy-PLAN simulation environment to assess the hourly feasibility of clean energy generation for power and freshwater production through desalination. Their research showed that increasing the RE share to 52% in Morocco by 2030 could lead to a 52.09% reduction in carbon emissions. Their empirical result highlights how RE may greatly enhance environmental quality, especially with technological advances.

Al-Ayouty [27] studied 18 MENA nations plus Turkey to scrutinize the interconnection between economic complexity and pollutant emissions from 1990 to 2020. The author employed panel data econometric techniques based on the Fully Modified Ordinary Least Squares (FMLOS) and Granger causality tests. The estimation findings indicated that green energy adoption demonstrates a substantial inverse relationship with CO₂ output, suggesting a significant amelioration in environmental quality. In addition, the econometric estimations validate the Environmental Kuznets Curve concept across the MENA and Turkey. This study’s key policy implication is that Turkey should acknowledge the need to pair EG and industrialization with a shift from traditional to knowledge-based economies. Albaker et al. [5] analyzed the effects of shifting to sustainable energy on EG and the emission of greenhouse gases in the MENA region between 1997 and 2021. Advanced econometric techniques (e.g., Method of Moments Quantile Regression) were used. They concluded that RE use adds significantly to economic expansion and reduces CO₂ emissions, implying more investment in renewable technologies.

Alharthi et al. [28] investigated the main factors of CO₂ evolution in MENA countries that adopted empirical aspects of the EKC structure by employing econometric quantile techniques, using data set from 1990-2015 on real per capita income, CO₂ emissions, fossil and green energy sources, and urbanization. The results of their study demonstrated that carbon discharge decreases drastically with the use of sustainable energy, which is observed to be more effective at higher quantiles. This means, in turn, that an expansion in the size and overall real income of economies makes sustainable energy relatively more effective at reducing environmental harm.

On the other hand, some other research reveals that RE consumption may not always translate into environmental improvement. In this context, we can mention the study by Kahia et al. [10] for 24 countries in the MENA area from 1980 to 2012 to analyze the potential impact of sustainable energy use on CO₂ evolution, EG, and financial development. To do so, the authors used the Panel Vector autoregression technique. Their findings revealed that RE use had little influence, indicating that these nations’ energy sectors are immature. The results also showed that the examined countries should strengthen the RE industry by facilitating bank loans for green energy project investments to prevent environmental harm and boost EG.

Similarly, in their study, Bélaïd and Youssef [29] applied the ARDL cointegration technique to analyze the correlation between CO₂ emissions and renewable electricity usage in Algeria from 1980 to 2012. Their econometric findings indicate that renewable electricity usage shows no statistically significant effect on CO₂ emissions in the short or long term. However, their empirical results indicate that the utilization of renewable power has a Granger-causal relationship with long-term CO₂ emissions. Lin and Abudu [30] analyzed energy-saving and substitution policies and plans to lower CO₂ output in the electricity sector in the MENA region. Using data from 2000 to 2015, they found that RE intensity is too low to reduce CO₂ discharge effectively.

Moreover, non-renewable energy intensity remains high, contributing to poor energy efficiency at just 40%. Consequently, current production methods and policies, like subsidies for non-renewable energy, block significant CO₂ reduction. Thus, the region’s EG is tied to non-renewable energy use, meaning that cleaner energy and climate goals need substantial policy and technology changes.

2.3 The economic growth-renewable energy consumption nexus

The EG-RE consumption nexus has secured substantial attention in the literature due to its implications for sustainable development and policy planning. In this context, using panel data techniques, Bhattacharya et al. [31] examined the interconnection between RE usage and EG across 38 countries from 1991 to 2012. The estimation findings demonstrated a positive long-term elasticity between RE investments and EG in more than half of the sample, underscoring the global significance of RE in promoting sustainable development. In the same regard, Kayesh and Siddiqa [32] also examined the dynamic associations between RE consumption, natural gas consumption, and EG in Bangladesh from 1980 to 2018. The authors used a combination of the Autoregressive Distributed Lag (ARDL) model for long-term relationship analysis and the Vector Error Correction Model (VECM) for causal inference. Their results revealed a unidirectional long-run causality from RE usage to EG, indicating the key role in achieving Bangladesh’s sustainable development goals. In their work, Yildirim et al. [33] explored the U.S. economy’s RE consumption–EG nexus over a six-decade period from 1949 to 2010 using bootstrap-corrected and Toda-Yamamoto causality tests. Their empirical result indicated that while biomass-waste-derived energy positively influences GDP, other renewable sources exhibit weaker economic ties.

Similarly, Ocal and Aslan [34] examined the connection between RE consumption and EG in Turkey from 1990 to 2010, using ARDL bounds testing and Toda-Yamamoto causality. Their econometric results provided evidence from Turkey supporting the conservation hypothesis, which suggests that EG drives RE consumption rather than the reverse. These findings show that EG and RE consumption interact in a complex and context-specific manner and highlight the necessity of local and targeted energy policies.

Our review of the literature on the renewable energy-environmental pollution nexus in the MENA region shows that most studies focus on the growth-environment link in isolation, overlooking the dynamic relationship between renewable energy (RE) adoption and pollution. This gap limits understanding of how RE adoption and economic growth affect carbon emissions, especially in the MENA region, where RE sectors are still developing, and economic structures vary significantly. Addressing this relationship is crucial to identifying feedback mechanisms and guiding policy. By analyzing these dynamics, policymakers can better promote economic growth and environmental sustainability, while future research can provide stronger empirical evidence to improve models and predict the long-term effects of green energy transitions. Our study aims to fill this gap and offer actionable policy recommendations for the MENA region.

Our econometric analysis aims to explore how EG and RE use influence environmental pollution across a selection of Arab nations. In line with this objective, the initial step involved testing the hypothesis that all variables exhibit cross-section dependence. Then, in the second step, we examined the time series stationarity and determined the integration order of each variable in our study. The third step involves investigating a long-run cointegrated connection among the selected variables. In the fourth and final step, we applied two estimators: the MG estimator, formulated by Pesaran and Smith [40], and the PMG estimator, introduced by Pesaran [41].

4.1 Testing cross-sectional dependencies and slope homogeneity

The study employed several tests for cross-sectional dependence to detect possible dependencies among cross-sections. These include the formulae formulated by Pesaran [35], the Bias-corrected scaled LM test, the Breusch-Pagan LM test and the Pesaran scaled LM test. The null hypothesis of all these tests assumes cross-sectional independence. Table 3 summarizes the test results. We can notice that at a 1% significance level, the findings indicate that all variables in the Panel significantly rejected the null hypothesis. The test results substantiate the cross-sectional dependence among all variables, suggesting that the data exhibit interdependencies across various panel units. 

Table 4 provides the Pesaran [41] proposed cross-sectional dependence test findings. It indicates that the null hypothesis of cross-sectional independence is statistically rejected at the 5% level and 1% significance levels for the fixed and random effect models, respectively. Furthermore, the adjusted LM test for cross-sectional dependence is significant at the 1% level for both the fixed effect (FE) and ransom effect (RE) models.

Table 2. Descriptive statistics overview

Note: The symbols (***), (**) and (*) indicate the statistical thresholds of 0.01, 0.05, and 0.1 in succession

Table 3. Cross-sectional dependence tests

Note: The symbols (***), (**), and (*) indicate hypothesis H₀: Cross-sectional independence is not accepted at statistical thresholds of 0.01, 0.05, and 0.1 in succession

Table 4. Model comparisons for cross-sectional dependence

Note: The symbols (***), (**), and (*) indicate hypothesis H₀: Cross-sectional independence is not accepted at statistical thresholds of 0.01, 0.05, and 0.1 in succession

Table 5. Slope homogeneity test

Note: The symbols (***), (**), and (*) indicate hypothesis H₀: slope homogeneity is not accepted at statistical thresholds of 0.01, 0.05, and 0.1 in succession

In addition, we applied the slope homogeneity test introduced by Pesaran and Yamagata [36].

The results outlined in Table 5 show that the null hypothesis of slope homogeneity is rejected with a significance level of 1% for the statistics. $\Delta_{a d j}^{\sim}$ and $\Delta_{a d j}^{\sim}$. Therefore, it is imperative to consider slope heterogeneity and cross-sectional dependence to prevent estimate bias.

4.2 Panel stationarity tests

Given that our empirical findings revealed interdependencies among cross-sections and heterogeneous slopes across the examined variables, we employed an advanced second-generation methodology to assess stationarity. Specifically, we applied the Cross-Sectional Im, Pesaran, and Shin (CIPS) and the Cross-Sectional Augmented Dickey-Fuller (CADF) tests. The results, presented in Table 6, show that the CIPS and CADF tests confirm the presence of a unit root in all panels, indicating that the variables are not stationary at level, except for lnRGDP, which is stationary, as the P-values for both tests are less than 0.01. However, the first-generation unit root test, such as Maddala and Wu’s Fisher test [42], reveals that all variables are initially non-stationary but become stationary after first-order differencing. We prioritized the findings of the CIPS test due to the model’s cross-sectional dependence. Based on this approach, we conclude that the variables lnCO₂ and lnREC are integrated of the first order, I(1), while lnRGDP is I(0).

Table 6. Panel unite root tests

Note: The symbols (***), (**), and (*) indicate H₀ = all panels contain unit roots is not accepted at statistical thresholds of 0.01, 0.05, and 0.1 in succession

4.3 Cointegration tests

After we identified that the variables display mixed integration orders, ranging from I(0) to I(1), we explored the presence of cointegrated relationships among them. To this end, we employed three cointegration tests: the Pedroni test [43], the Kao test [44], and the Westerlund test [45]. We specifically included the Westerlund test alongside the widely used Pedroni and Kao tests due to its ability to accommodate cross-sectional dependencies, which is crucial for accurate inference in our panel data where such interdependencies are present. The null hypothesis of all these tests posits no long-run cointegrated connection among the variables.

Table 7. Cointegration tests

Note: The symbols (***), (**), and (*) indicate H₀: No cointegrated relationship is rejected at statistical thresholds of 0.01, 0.05, and 0.1 in succession

The cointegration test results are shown in Table 7. The empirical results illustrate the rejection of the null hypothesis at a 1% significance level, approximately for the most applied tests. Exceptions include the ADF statistic of the Kao test, where the null hypothesis is rejected at a 10% level, the variance ratio statistic of the Westerlund test, and the Panel rho-statistic and Group rho-statistic of the Pedroni test, which are significant at the 5% level. The empirical outcomes support the presence of a long-run cointegrated nexus among the variables, indicating a long-term equilibrium and causal connection between RE usage, economic expansion, and CO₂ discharge. The discovery of these enduring connections helps us achieve the goal of this study. It enables us to assess how green energy utilization and economic expansion affect the evolution of CO₂ emissions in the long and short run across this group of six MENA countries.

4.4 Model estimations

After assessing the stationarity results of the panel data series, we found that the variables demonstrate zero-order integration I(0), and first-order integration I(1). Given the different integration orders, we applied the Panel Autoregressive Distributed Lag (ARDL) method, which is suitable for such mixed-order series, to explore the dynamic interaction among the variables. To explore and estimate the effects of RE usage and economic expansion on CO₂ trend emissions, we applied the ARDL model with the configuration ARDL (1,3,3). This model used the Akaike Information Criterion (AIC) and the Schwarz Bayesian Criterion (SBC) as guiding factors. We also included robust estimation methods such as the MG estimator, formulated by Pesaran and Smith [40], and the PMG estimator, introduced by Pesaran [41]. This methodology provides an extensive framework for studying the dynamic interactions between green energy usage, economic expansion, and CO₂ discharge across different groups within the panel data context. The MG estimator in panel data models adapts to heterogeneity across units by allowing for individual-specific effects while assuming homogeneous time-specific effects across all cross-sections. This method involves estimating models for each unit individually and then averaging these estimates to drive the mean group estimating.

However, the PMG estimator is utilized to assess the long-run relationships between the variables. In general, the PMG estimator outperforms the MG estimator in terms of efficiency, provided that the constraints on the long-term coefficient remain valid. It involves averaging and pooling the coefficients whereby homogeneity is imposed on the long-run coefficients across the cross-sections while allowing heterogeneity in the error variance intercepts and short-run coefficients [46].

To determine whether the MG or PMG estimator was more suitable for our panel data model, we based our analysis on the Hausman test. The null hypothesis considers that the PMG estimator, which assumes long-term coefficient homogeneity, is preferable to the MG estimator. In other words, the null hypothesis evaluates the validity of long-term homogeneity restrictions among the countries in our sample. The findings presented in Table 8 show that the null hypothesis could not be upheld, indicating that the MG estimator, which accounts for heterogeneity in both short-run and long-run coefficients, provides a more efficient fit for our data.

Table 8. Panel long and short run coefficients: Mean Group Estimator (MG)

Note: The symbols (***), (**) and (*) indicate the statistical thresholds of 0.01, 0.05, and 0.1 in succession

The choice of MG and PMG estimators provides critical insights into the dynamic interactions among RE consumption, EG, and CO₂ emissions. While the PMG estimator assumes homogeneity of long-run coefficients across countries, the MG estimator allows for heterogeneity in both short- and long-run coefficients, which proves more suitable for this study’s panel data, which includes countries with diverse economic structures and RE adoption rates. The Mean Group (MG) estimator accommodates parameter heterogeneity across countries, yielding consistent country-specific coefficients that capture individual slope parameters. This estimation approach reveals significant cross-country parameter heterogeneity in the relationships between EG, RE consumption, and environmental indicators. The robust performance of the MG estimator in this context demonstrates its relevance for studies involving heterogeneous panels.

Consequently, the MG estimator is appropriate for analyzing the short-term and long-term interactions between CO₂ emissions, RE usage, and economic expansion in the six Arab countries studied. Table 8 displays the short-run and long-run MG estimator outcomes for the model coefficients. Furthermore, it estimates the error correction term (ECT), which measures the speed at which the variables converge towards their long-term equilibrium. The ECT shows a negative value and reaches statistical significance at 1%, suggesting a fast return to the long-term equilibrium for the variables. This indicates a robust long-run connection between the variables studied. The estimated value of the error correction coefficient is -0.4, suggesting that approximately 40% of any deviation from the long-term equilibrium is corrected in each period. Hence, this significant adjustment speed verifies a stable long-run connection tying CO₂ emissions, RE use, and economic expansion.

We can notice from Table 8 that the long-term coefficients’ P-values of lnREC and lnRGDP are less than 0.01, which means that these coefficients are statistically acceptable at 1%. The empirical evidence supports a negative long-term link between RE use and pollutant (CO₂) emissions within this group of Arab countries. A 1% rise in RE consumption corresponds to a 0.13% decrease in CO₂ emissions. Additionally, the results demonstrate a robust and significant positive long-term nexus tying CO₂ emissions and EG, where a 1% expansion in EG causes a 0.84% augmentation in CO₂ emissions. However, in the short term, the coefficients of RE consumption and EG lack statistical significance, indicating that short-term fluctuations in these variables do not affect environmental pollution in the examined Arab countries.