Asymmetric Moderating Role of Geopolitical Risk in the Relationship Between Oil Rents and CO₂ Emissions in Saudi Arabia: An NARDL Approach

Abstract

Oil is a major source of income and emissions in the Saudi economy. Thus, this study examines the symmetrical and asymmetrical impacts of oil rents (ORs) on CO₂ emissions using data from 1970 to 2024. For this purpose, the Nonlinear Autoregressive Distributed Lag (NARDL) model is applied, while the conventional ARDL model is used as a baseline model. In addition, the moderating effect of geopolitical risk (GPR) is also tested in the association between OR and emissions. The Environmental Kuznets Curve is validated in both the long run and the short run. Moreover, OR is found to be a major driver of emissions, and positive shocks amplify emissions more than negative shocks reduce them. GPR has a negative relationship with emissions, and the impact of positive shocks in GPR is found to be greater than the impact of negative shocks in GPR. However, the interaction between OR and GPR shows that simultaneous increases in both factors exacerbate emissions, whereas the effect of negative shocks in this interaction is insignificant. Thus, asymmetry is corroborated in all investigated relationships. Moreover, the Wald tests also confirm significant asymmetric relationships. The findings suggest reducing oil dependence and adopting GPR-sensitive planning to mitigate the environmental impacts of the oil sector in line with Vision 2030 and the Sustainable Development Goals (SDGs).

1. Introduction

The oil sector is a major source of income for the Saudi economy, but it also poses serious environmental threats because of extensive Greenhouse Gas (GHG) emissions generated throughout the stages of oil extraction, production, and consumption. Moreover, oil remains a primary energy source for electricity generation, transportation, and other commercial and residential energy needs in Saudi Arabia. Therefore, the combustion of oil for energy production significantly contributes to GHG emissions and environmental pollution. In recent years, Saudi Arabia has ranked among the world’s top 10 polluting economies (Worldometer, 2025). In addition, oil rents (ORs) constitute a major source of export earnings and government revenues, which are used to finance urban development, infrastructure projects, and social programs that may generate additional environmental pressure through demand-side effects. OR is the net value of crude oil production after deduction of the cost of production, which is expressed as a percentage of Gross Domestic Product (GDP) (World Bank, 2026). Alongside these environmental concerns, Saudi Arabia is located in the Middle East, a region characterized by high geopolitical risk (GPR) arising from conflicts, wars, political tensions, and policy uncertainty. From an environmental perspective, higher GPR may reduce economic activities and pollution levels because uncertainty can weaken investment and production activities. However, GPR may also discourage investment in renewable energy infrastructure and green transition projects, which are key objectives under Saudi Vision 2030 and are also aligned with the SDGs.

This study follows the Prospect Theory of Kahneman and Tversky (1979), which advocates that decision-makers respond differently to gains and losses under conditions of risk and uncertainty. Accordingly, policymakers in Saudi Arabia may react differently to increases and decreases in oil rents, leading to asymmetric effects of oil rents on CO₂ emissions. For instance, during periods of high GPR, the government may prioritize oil rent generation and economic stability over environmental sustainability concerns, which can strengthen the positive association between OR and emissions. Conversely, during periods of low geopolitical risk, policymakers may adopt a more flexible and sustainability-oriented approach, which may weaken the relationship between oil rents and environmental degradation. Moreover, the government may allocate a larger share of oil revenues toward renewable energy development and green transition projects under stable geopolitical conditions. From a global perspective, GPR can also influence international oil prices, as observed during recent Iran–US geopolitical tensions. Conflicts and wars may threaten global oil supply chains, leading to higher oil prices in international markets. Under relatively stable production and export conditions, rising oil prices can increase Saudi Arabia’s oil revenues and oil rents. However, heightened geopolitical risk may simultaneously slow global economic activities and reduce energy demand because of rising uncertainty, which may eventually place downward pressure on oil prices. In another extreme scenario, attacks on oil facilities in Saudi Arabia may disrupt oil production and exports, thereby reducing oil production capacity and oil rents even during periods of high oil prices. Furthermore, geopolitical risk can increase volatility in global oil markets, creating uncertainty for both the government and private sector in undertaking long-term investments in renewables.

In the OR, GPR, and the environment nexus, the literature has worked separately on the connection between OR and pollution (Lin et al., 2024; W. Wang et al., 2023) and GPR and pollution (Anser et al., 2021; Ding et al., 2023). However, the literature is scant in exploring how GPR may change the impact of OR on emissions (W. Zhao et al., 2021). Moreover, it is not necessary that the effects of OR and GPR on emissions are symmetrical. For instance, increasing oil prices and rents lead to higher oil production, which increases emissions. However, it can be invested in renewable energy infrastructure, which can reduce emissions. So, the net effect of increasing OR can be uncertain, which depends on government decisions and economic conditions. Alternatively, a fall in oil prices may motivate oil producers to produce more to sustain OR, which may generate more emissions. However, a fall in oil prices may also reduce oil production and emissions. But lower oil revenues can limit investment in environmental projects. Thus, the effects of positive and negative shocks to oil prices and rents could have asymmetrical effects.

Similarly, asymmetry can be expected in the GPR–emissions nexus. For instance, a rising GPR would reduce economic activities, oil production, and emissions in the short run. However, a persistently high GPR may shift the government’s focus toward energy security and OR, which can increase emissions in the long run. Conversely, a decreasing GPR may have a milder impact on emissions. For instance, reduced GPR can improve stability in government behavior toward long-term planning, green investment, and stronger environmental policies, which may reduce emissions gradually. Thus, the response of the environment to GPR is not symmetrical or instantaneous under both rising and falling GPR conditions. Likewise, the moderating effect of GPR can also be asymmetrical. Increasing GPR may strengthen the positive relationship between OR and emissions, as governments often prioritize securing revenues and maintaining economic stability. Consequently, oil production and emissions may increase, and environmental policies may be weakened. However, declining GPR may result in a weaker or even inverse relationship between OR and emissions. For instance, OR may be invested in renewable energy production under low GPR conditions.

The above discussion highlights the importance of GPR and OR in determining environmental quality in Saudi Arabia. Several Saudi-based studies have examined the role of GPR in explaining economic diversification (Sweidan & Elbargathi, 2023), economic complexity (Aloui et al., 2024), trade performance (Islam & Islam, 2026), international competitiveness (Su et al., 2019; Razek & McQuinn, 2021), insurance demand (Hemrit, 2021), and climate-related variables (Dhifaoui et al., 2023). Some studies have also explored the moderating effect of GPR in the nexus between FDI, trade, and CO₂ emissions (Alsubaie et al., 2025), as well as between OR and economic development (Sweidan & Elbargathi, 2022). However, no study has examined the moderating effect of GPR on the connection between OR and emissions. In addition, the simultaneous examination of OR, economic growth, GPR, and emissions has also not been investigated in resource-dependent economies. This study is therefore motivated to estimate the direct impact of OR and GPR on emissions in the Saudi economy. Moreover, it also examines whether GPR moderates the effect of OR on emissions. Finally, the possibility of asymmetry is tested across all hypothesized relationships. Accordingly, a nonlinear ARDL (NARDL) model is applied to data from 1970 to 2024 in the oil-dependent Saudi economy, which is still missing in the existing Saudi literature. Overall, the following research questions are examined:

• Does OR have a symmetrical or asymmetrical impact on CO₂ emissions in Saudi Arabia?
• Does GPR directly affect CO₂ emissions symmetrically or asymmetrically?
• Does GPR moderate the relationship between oil rents and CO₂ emissions?
• Is the Environmental Kuznets Curve (EKC) valid in Saudi Arabia over the period 1970–2024 in the presence of GPR, OR, and their interaction in the model?

2. Literature Review

2.1. Negative Environmental Effects of GPR

GPR could have a different impact on the environment. The literature suggests that GPR can disrupt environmental regulatory frameworks. Many studies have investigated the GPR and emissions relationship in BRICS economies. Anser et al. (2021) examined the impact of GPR on CO₂ emissions in BRICS countries from 1985 to 2017 and found that a 1% increase in GPR raised CO₂ emissions by 13%, indicating disruption in long-term environmental planning and delayed green investments. Khan et al. (2023) re-examined BRICS countries over the period 2000–2021 using a fully modified bias-corrected technique and found that both GPR and militarization increased emissions and energy usage. In addition, bidirectional relationships between GPR and environmental degradation were also reported. Using similar econometric techniques, Guo et al. (2023) included nuclear energy efficiency in the analysis of BRICS economies and found that it reduced carbon emissions; however, the effects of GPR and Economic Policy Uncertainty (EPU) were found to be ineffective in reducing carbon emissions.

Using Cross-Sectional Dependence (CSD) techniques, Uddin et al. (2023) re-investigated BRICS countries by including Institutional Quality (IQ) indicators in the analysis. The authors found that IQ and Foreign Direct Investment (FDI) condensed emissions; however, GPR positively affected CO₂ emissions. In a CSD framework, Cui et al. (2024) re-examined BRICS countries from 1992 to 2021 by incorporating both aggregated IQ and EPU in the analysis and revealed that IQ and energy productivity mitigated CO₂ emissions, whereas GPR and EPU increased pollution. In a sectoral analysis, Kartal et al. (2025) analyzed and confirmed the heterogeneous influence of GPR on sectoral emissions in BRICS countries over the period 2000 to 2022. In a quantile-based approach, Syed et al. (2022) investigated BRICS countries and Turkey and reported a heterogeneous environmental effect of GPR; for instance, GPR increased emissions at earlier quantiles but reduced emissions at higher quantiles. Thus, the impact of GPR depends on a country’s existing emission levels.

Qamruzzaman et al. (2025) re-examined BRICS economies over the period 2004 to 2022 and found that green innovation and renewable energy consumption (REC) improved environmental quality. Nevertheless, GPR and unplanned urbanization increased emissions. Moreover, feedback effects were observed between GPR, urbanization, and environmental degradation. Luo and Sun (2024) indicated that GPR accelerated emissions; however, environmental policy stringency weakened the impact of GPR. In a panel of 38 countries with different development levels, L. Chen et al. (2024) used data from 1990 to 2019 and reported that GPR amplified CO₂ emissions inequality. Kolati and Raghutla (2025) examined leading economies from 1990 to 2022 and substantiated that REC, fintech development, and industrialization mitigated CO₂ emissions, while GPR intensified them. Sanusi et al. (2025) studied China, India, and the USA from 1990 to 2019 and confirmed that GPR increased CO₂ emissions. In a causality analysis for China, K.-H. Wang et al. (2022) reported feedback between GPR and CO₂ emissions. Similarly, for South Africa, Lawal et al. (2023) also confirmed feedback causality between GPR and CO₂ emissions.

Using CSD techniques, Zhang et al. (2024) explored G7 nations from 1990 to 2021 and substantiated that energy consumption and GPR negatively affected environmental quality. Shanyazov et al. (2026) re-explored the G7 from 1990 to 2022 and found that GPR increased emissions. Moreover, policy uncertainties also contributed to CO₂ emissions. Kisswani et al. (2024) explored Southeast Asia from 1990 to 2020 and found that climate policy uncertainty reduced CO₂ emissions. However, GPR and Gross Domestic Product (GDP) increased emissions. Liza et al. (2024) probed the G20 and reported that green finance reduced pollution. Nevertheless, GPR positively impacted emissions. Shen et al. (2024) scrutinized 156 economies from 2000 to 2022 and reported that increasing GPR and migration positively contributed to emissions. Flouros et al. (2022) evaluated 171 economies and reported that GPR reduced green investments.

2.2. Positive Environmental Effects of GPR

The literature has also corroborated the positive environmental effects of GPR through reducing pollution, as GPR can reduce economic activities and emissions. For instance, Hashmi et al. (2022) utilized a global dataset and substantiated that GPR condensed emissions globally due to an immediate economic slowdown. Nevertheless, GPR raised emissions in the long run. Ma et al. (2022) found that GPR reduced emissions in the short run and increased them in the long-term. Moreover, the EKC was validated. Husnain et al. (2022) investigated E7 and corroborated that GPR reduced CO₂ emissions. Kızılkaya et al. (2024) evaluated Turkey from 1985 to 2019 and reported that increases in GPR reduced CO₂ emissions. Villanthenkodath and Pal (2024) investigated India from 1990 to 2019 and found that GPR improved environmental quality. Pata et al. (2023) analyzed sectoral CO₂ emissions from 1997 to 2022 and validated that GPR and EPU reduced CO₂ emissions in most sectors because of production disruptions. Still, GPR increased emissions in the transportation sector.

Ulussever et al. (2025) investigated GCC from 2000 to 2021 and corroborated that GPR showed mixed impacts on CO₂ emissions in different quantiles. Nawaz et al. (2023) examined Italy from 1997 to 2019 and found that religious tourist arrivals and GPR reduced CO₂ emissions. However, FDI and transportation contributed to CO₂ emissions. Rashid and Gopinathan (2023) investigated India and found that GPR reduced CH₄ and N₂O emissions, substantiating the EKC. Saadaoui et al. (2024) explored Turkey from 1985 to 2021. GPR, hydroelectricity, and FDI reduced emissions. Zeng et al. (2025) investigated Mexico, Indonesia, Nigeria, and Turkey from 1990 to 2020 and found that REC and EPS increased emissions. However, GPR and innovation reduced CO₂ emissions. Yuen and Yuen (2024) scrutinized the role of GPR on R&D and found that financially capable governments increased investments in renewable R&D to safeguard energy supply during high GPR. Alsagr and van Hemmen (2021) assessed emerging markets from 1996 to 2015 and reported that GPR raised REC. Rasoulinezhad et al. (2020) analyzed Russia from 1993 to 2018 and found that GDP, population, and inflation reduced energy transition. However, CO₂ emissions, GPR, exchange rate, and financial openness promoted energy transition.

2.3. The Resource–Emissions–GPR Nexus

The reliance on resource rents, particularly from fossil fuels, may exacerbate emissions in any resource-abundant nation. In an extension of the resource curse theory, studies have investigated the ecological effects of Natural Resource Rents (NRRs). Iqbal et al. (2025) probed 36 economies from 2000 to 2021 and reported that artificial intelligence, GPR, and NRR amplified emissions. However, REC and IQ reduced emissions. In the mixed sample of 36 economies from 2000 to 2020, Lin et al. (2024) evaluated and confirmed that NRR, productive capacity index, and GPR positively affected CO₂ emissions. Nevertheless, globalization and REC reduced CO₂ emissions. Bello and Hassan (2024) probed the influence of GPR on REC and found that a 1% increase in GPR reduced REC by 0.22% in OECD economies from 1970 to 2022. Moreover, CO₂ emissions and NRR adversely affected REC. However, economic growth and globalization promoted REC.

Ding et al. (2023) analyzed 25 OECD and substantiated that GPR increased CO₂ emissions, which were primarily driven by risks related to mineral resource extraction. Moreover, GPR promoted fuel imports, which also contributed to emissions. In quantile analysis, Du and Wang (2023) considered China from 1995 to 2020. GPR, agricultural development, and NRR raised CO₂ emissions in the upper quantiles. However, green financing reduced them in all quantiles. W. Wang et al. (2023) examined China from 1988 to 2021 and reported that NRR increased CO₂ emissions. However, GPR in China significantly contributed to lowering emissions. Z. Zhao et al. (2023) scrutinized 20 OECD nations from 1970 to 2019 and found that both GPR and NRR reduced REC. Z. Chen et al. (2025) examined 38 nations from 1990 to 2021 and validated that NRR, GPR, and REC improved the energy security index.

2.4. Asymmetrical Analysis and GPR as a Moderator

In the methodological contribution, the literature has investigated the moderating effect of GPR. L. Chen et al. (2023) investigated 38 industrialized countries from 1970 to 2021 and substantiated that GPR and NRR amplified CO₂. Moreover, the moderation of GPR further increased the effect of NRR on CO₂ emissions. Thus, GPR had direct adverse environmental effects and also had indirect adverse environmental effects through NRR’s interaction. Kartal et al. (2024) analyzed daily data from 2019 to 2024 and established that REC condensed CO₂ emissions. Nevertheless, GPR decreased this effect.

By using System GMM, Y. Zhao et al. (2024) examined BRICS by using data from the period 1990–2018. The authors reported that GPR directly raised CO₂ emissions. Nevertheless, the influence of the interaction of green technology and GPR reduced CO₂ emissions. Thus, green technology helped GPR to reduce emissions. Fatima et al. (2026b) augmented the relationship between GPR and emissions by adding EPS in BRICS economies from 1990 to 2020 and substantiated that EPS and green technology innovation directly condensed emissions. Moreover, their interaction effect also helped reduce emissions. However, GPR increased emissions directly. Fatima et al. (2026a) assessed the moderating role of EPS in 13 EU countries from 1990 to 2022 and found that the interaction of EPS with GPR, innovation, and energy transition promoted environmental sustainability. In the asymmetric analysis of BRICS countries from 1985 to 2019, W. Zhao et al. (2021) reported that GPR had negative effects on emissions in Russia, and falling GPR reduced emissions in India and China. Moreover, the positive coefficient of GPR was found in Russia. Thus, GPR asymmetrically and differently affected the energy and environmental variables in different BRICS economies.

2.5. The Saudi GPR Literature and a Research Gap

The recent literature on Saudi Arabia has examined the role of GPR in macroeconomic performance. However, limited attention has been paid to its implications for environmental degradation. For instance, Islam and Islam (2026) investigated the effects of GPR on Saudi metal exports and found that GPR negatively affected trade performance. Aloui et al. (2024) explored the impact of GPR, uncertainty, and oil price volatility on economic complexity and found that GPR and oil price volatility positively influenced economic complexity. Moreover, Sweidan and Elbargathi (2023) showed that oil prices and GPR hindered economic diversification. Guizani (2024) and Guizani et al. (2023) demonstrated that GPR and EPU significantly influenced firms’ financing and cash-holding decisions. Hemrit (2021) further revealed that GPR and EPU negatively affected insurance demand. Some studies emphasized the importance of geopolitical conditions in shaping oil prices, liquidity, and international competitiveness (Su et al., 2019; Razek & McQuinn, 2021). Moreover, Dhifaoui et al. (2023) identified dynamic causal links between GPR, total precipitation, relative humidity, temperature, and wind speed and direction. Some studies also tested the moderating effect of GPR in the association between FDI, trade, and CO₂ emissions (Alsubaie et al., 2025) and OR and economic development (Sweidan & Elbargathi, 2022). However, no study has specifically examined whether geopolitical risk moderates the relationship between OR and environmental degradation. Moreover, most studies relied on symmetric modeling approaches. However, the symmetric assumption may be unrealistic in an oil-dependent economy, where increases and decreases in OR or GPR may generate different environmental responses. In addition, no study has integrated the moderating effect of Saudi GPR in the OR–emissions nexus. Therefore, this research contributes to the literature by examining the asymmetric moderating effect of Saudi GPR in the association between OR and CO₂ emissions by using the NARDL approach. Particularly, GPR can shape oil prices and rents as per the recent US–Iran conflicts, which can consequently affect carbon emissions. Considering a gap in the Saudi literature, this paper augments the literature by probing these relationships. To add novelty to the study, all direct and moderating effects are tested by using NARDL.

3. Methodology

GDP growth is a basic component of the CO₂ emissions model, and this relationship is nonlinear as per the EKC theory (Grossman & Krueger, 1991). Thus, early GDP growth may increase emissions, which can be reduced with further GDP growth after a threshold point. OR can play a significant role in the EKC framework as OR is a major part of the GDP in Saudi Arabia, which is the value of crude oil production net of production costs (World Bank, 2026). The resource curse literature argues that resource wealth may distort economic incentives, which can affect institutional capacity and disturb environmental quality in oil-dependent economies. Saudi Arabia is a rentier state, and oil rents have been invested in its modernization without sufficient investment in renewable markets. The literature has signified a positive connection between OR and emissions in oil-dependent economies (Lin et al., 2024; W. Wang et al., 2023). On the positive long-term aspects, OR can be invested in renewable markets and infrastructure. However, GPR shapes investment decisions in a different time horizon. For instance, an increase in GPR may reduce long-term environmental investments, and short-term OR maximization can be preferred over long-run sustainability concerns (Anser et al., 2021). Therefore, GPR moderates the impact of OR on emissions (L. Chen et al., 2023). However, combining economic growth (GDP per capita) with OR (OR percentage of GDP) may raise multicollinearity and endogeneity concerns in the model because a significant amount of GDP depends on OR in Saudi Arabia. To accommodate this issue, the Autoregressive Distributed Lag (ARDL) framework of Pesaran et al. (2001) is applied, which is specifically designed to accommodate these issues by allowing lagged differenced variables and provides consistent estimates regardless of potential endogeneity among regressors. Thus, the model is hypothesized as:

$${\mathrm{C}\mathrm{O}2}_t=f({\mathrm{Y}}_t,Y_t^2,{\mathrm{O}\mathrm{R}}_t,{\mathrm{G}\mathrm{P}\mathrm{R}}_t,{\mathrm{G}\mathrm{P}\mathrm{R}}_t\times {\mathrm{O}\mathrm{R}}_t)$$

(1)

CO_2t is the natural log of carbon dioxide (CO₂) emissions excluding Land Use, Land-Use Change, and Forestry (LULUCF), which are based on the agriculture, energy, waste, and industrial sectors. This measure excludes GHG fluxes caused by LULUCF (World Bank, 2026). Y_t is the natural log of GDP per capita (constant LCU). Moreover, the square term Y_t² is added, which can be used to test the EKC in nonlinear form. OR_t shows the natural logarithm of the OR percentage of GDP. These data series are sourced from the World Bank (2026). GPR_t is converted from a monthly series to an annual series of the Saudi GPR index, and this data has been taken from the Economic Policy Uncertainty Database (2026) as per the methodology of Caldara and Iacoviello (2022). t represents the time period from 1970 to 2024.

In Equation (1), potential endogeneity and reverse causality between CO₂ emissions and per capita GDP are also expected. Therefore, the study applies the ARDL methodology of Pesaran et al. (2001), which partially addresses this issue through its dynamic specification. For instance, the ARDL model incorporates lagged values of both the dependent and explanatory variables, which allows the current level of CO₂ emissions to depend on past realizations of economic growth and other regressors, thereby mitigating the potential problem of reverse causality and simultaneity bias between economic growth and CO₂ emissions. Moreover, ARDL reduces omitted variable bias through appropriate lag selection, and the error correction representation further captures the speed of adjustment toward long-run equilibrium, which helps limit contemporaneous feedback effects between GDP and CO₂ emissions. However, the series in Equation (1) should be tested for the unit root problem to proceed with the ARDL methodology. The unit root is tested by the four test statistics proposed by Ng and Perron (2001) in the following way:

$${MZ}_a^d={\left[{\displaystyle \frac{Y_T^d}{T}}\right]}^2/2K-f_0/2K$$

(2)

$${MSB}^d={\left[{\displaystyle \frac{k}{f_0}}\right]}^{1/2}$$

(3)

$${MZ}_t^d={MZ}_a^d·{MSB}^d$$

(4)

$${MPT}_T^d={[c}^2·K+{\displaystyle \frac{1-c}{T}}]·{\displaystyle \frac{Y_T^d}{f_0}}$$

(5)

Y_T^d is the detrended series, T is the sample size, K is the lag truncation parameter, f₀ is the spectral density estimator at zero frequency, and c is the local alternative parameter. The ARDL framework requires that all independent series should be stationary at the level or difference. However, the CO_2t must be stationary at its first difference. In the presence of these conditions, Equation (1) can be written in the following ARDL framework:

$$\begin{array}{l}{∆CO2}_t=a_{10}+a_{11}{CO2}_{t-1}+a_{12}{Y}_{t-1}+a_{13}{Y}_{t-1}^2+a_{14}{OR}_{t-1}+a_{15}{GPR}_{t-1}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+a_{16}{(GPR\times OR)}_{t-1}+{\sum }_{i=1}^j{a}_{17i}∆{CO2}_{t-i}+{\sum }_{i=0}^j{a}_{18i}{∆Y}_{t-i}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\sum }_{i=0}^j{a}_{19i}{Y}_{t-1}^2+{\sum }_{i=0}^j{a}_{20i}∆{OR}_{t-i}+{\sum }_{i=0}^j{a}_{21i}∆{GPR}_{t-i}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\sum }_{i=0}^j{a}_{22i}∆{(GPR\times OR)}_{t-i}+e_{1t}\end{array}$$

(6)

Δ is the first difference operator, and a₁₀ is the intercept term. e_1t is the white noise error term. At first, Equation 6 will be tested for a Bound test with a null hypothesis of (H₀: a₁₁ = a₁₂ = a₁₃ = a₁₄ = a₁₅ = a₁₆ = 0). If the F-value of the Wald test exceeds the upper bound critical value, then H₀ can be rejected, which will corroborate the cointegration. Later, a_12, a_13, a_14, a_15, a_16, and a₁₇ will be normalized by a₁₁, which will assess the long run effects. j represents the optimal lag length selected through AIC, with the maximum lag order restricted to 2, with the help of the following formula:

AIC = −2ln(L) + 2k

(7)

L is the maximum value of the likelihood function, and k is the number of estimated parameters (including intercept and variance terms). The AIC approach is chosen as it balances model goodness-of-fit with model parsimony by penalizing excessive lag structures in Equation (6). Moreover, a maximum lag length of 2 is imposed to avoid over-parameterization and loss of degrees of freedom with the sample size of 55 observations. In the ARDL framework with 4 regressors (Y_t, Y_t², OR_t, and GPR_t) and interaction term (GPR×OR)_t, increasing lag length excessively can significantly reduce the available degrees of freedom, which may lead to inefficient and unstable parameter estimates with additional lagged variables. Thus, restricting the maximum lag to 2 ensures a balance between capturing sufficient dynamic behavior in the relationships among CO₂ emissions, economic growth, oil rents, and geopolitical risk, along with maintaining model parsimony. This choice also helps to minimize the risk of multicollinearity among lagged regressors and avoids overfitting. Thus, the imposed lag restriction enhances the robustness, efficiency, and interpretability of the estimated ARDL model. Later, short-run effects can be captured by the following Error Correction Model (ECM). The ECM is derived from Equation (6) by replacing the lagged-level variable with the Error Correction Term (ECT_t−1):

$$\begin{array}{l}{∆CO2}_t=g_{10}{ECT}_{1t-1}+{\sum }_{i=1}^j{g}_{11i}∆{CO2}_{t-i}+{\sum }_{i=0}^j{g}_{12i}{∆Y}_{t-i}+{\sum }_{i=0}^j{g}_{13i}{Y}_{t-1}^2\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\sum }_{i=0}^j{g}_{14i}∆{OR}_{t-i}+{\sum }_{i=0}^j{g}_{15i}∆{GPR}_{t-i}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\sum }_{i=0}^j{g}_{16i}∆{(GPR\times OR)}_{t-i}+e_{2t}\end{array}$$

(8)

In Equation (8), g₁₀ is the coefficient of the error correction term (ECT_1t−1). g_11i–g_15i are short-run effects and e_2t is the white noise error term. The short-run relationship can be substantiated with a negative and significant value of $g_{10}$. All other coefficients will capture the short-run effects. Equations (1), (6) and (7) are assumed to capture the symmetrical effects of all independent variables. However, the nexus between OR and emissions can be treated as asymmetrical as per Prospect Theory (Kahneman & Tversky, 1979). For instance, investment decisions in the case of rising OR can be taken more cautiously as per Prospect Theory, and a part of the gain from OR can be invested in diversification or cleaner technologies, instead of investing all OR gains into fossil-fuel expansion. Thus, emissions should not increase in the same proportion as the increase in OR. In the case of negative shocks in OR, policymakers may react more strongly to losses compared to gains, in fear of fiscal and income loss. Thus, this situation can increase the extraction of oil to maintain the country’s income and fiscal capacity, which can delay green investment (W. Zhao et al., 2021). Thus, positive and negative shocks in OR could have asymmetrical effects on emissions in Saudi Arabia. In a similar way, these shocks in GPR could have different direct effects on carbon emissions and can also indirectly and differently affect the impact of OR on emissions. To test this asymmetrical nexus, all series are converted into positive and negative cumulative series by following Shin et al. (2014), except for GDP per capita (taken in quadratic form to test the EKC). The variables are transformed in the following way:

$${GPRP}_t={\sum }_{j=1}^t{GPR}_j^{+}={\sum }_{j=1}^{t}max(∆{GPR}_j,0)$$

(9)

$${GPRN}_t={\sum }_{j=1}^t{GPR}_j^{-}={\sum }_{j=1}^{t}min(∆{GPR}_j,0)$$

(10)

$${ORP}_t={\sum }_{j=1}^t{OR}_j^{+}={\sum }_{j=1}^{t}max(∆{OR}_j,0)$$

(11)

$${ORN}_t={\sum }_{j=1}^t{OR}_j^{-}={\sum }_{j=1}^{t}min(∆{OR}_j,0)$$

(12)

$$(GPR\ast OR)P_t={\sum }_{j=1}^t{(GPR\ast OR)}_j^{+}={\sum }_{j=1}^{t}max(∆{GPR\times OR)}_j,0)$$

(13)

$$(GPR\ast OR)N_t={\sum }_{j=1}^t{(GPR\ast OR)}_j^{-}={\sum }_{j=1}^{t}min(∆{GPR\times OR)}_j,0)$$

(14)

Equations (9), (11) and (13) estimate positive shocks of GPR_t, OR_t, and (GPR_t×OR_t). Equations (10), (12) and (14) estimate negative shocks of GPR_t, OR_t, and (GPR_t×OR_t). Considering the positive and negative shock variables, the symmetrical model in Equation (1) is written in the following asymmetrical form:

$${\mathrm{C}\mathrm{O}2}_t=f[{\mathrm{Y}}_t,Y_t^2,{\mathrm{G}\mathrm{P}\mathrm{R}\mathrm{P}}_t,{\mathrm{G}\mathrm{P}\mathrm{R}\mathrm{N}}_t,{\mathrm{O}\mathrm{R}\mathrm{P}}_t,{\mathrm{O}\mathrm{R}\mathrm{N}}_{t,}\left(GPR\times \mathrm{O}\mathrm{R}\right){\mathrm{P}}_t,\left(GPR\times \mathrm{O}\mathrm{R}\right){\mathrm{N}}_t]$$

(15)

The linear ARDL in Equation (6) and ECM in Equation (8) are also taken into asymmetrical forms as follows:

$$\begin{array}{l}{∆CO2}_t={b_{10}+b_{11}{CO2}_{t-1}+b_{12}{Y}_{t-1}+b_{13}{Y}_{t-1}^2+b}_{14}{ORP}_{t-1}+b_{15}{ORN}_{t-1}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+b_{16}{GPRP}_{t-1}+b_{17}{GPRN}_{t-1}+b_{18}\left(GPR\times OR\right)P_{t-1}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+b_{19}\left(GPR\times OR\right)N_{t-1}+{\sum }_{i=1}^j{b}_{20i}∆{CO2}_{t-i}+{\sum }_{i=0}^j{b}_{21i}{∆Y}_{t-i}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\sum }_{i=0}^j{b}_{22i}{Y}_{t-1}^2+{\sum }_{i=0}^j{b}_{23i}∆{ORP}_{t-i}+{\sum }_{i=0}^j{b}_{24i}∆{ORN}_{t-i}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\sum }_{i=0}^j{b}_{25i}∆{GPRP}_{t-i}+{\sum }_{i=0}^j{b}_{26i}∆{GPRN}_{t-i}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\sum }_{i=0}^j{b}_{27i}∆{\left(GPR\times OR\right)P}_{t-i}+{\sum }_{i=0}^j{b}_{28i}∆{\left(GPR\times OR\right)N}_{t-i}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+e_{3t}\end{array}$$

(16)

$$\begin{array}{l}{∆CO2}_t=g_{20}{ECT}_{2t-1}+{\sum }_{i=1}^j{g}_{23i}∆{CO2}_{t-i}+{\sum }_{i=0}^j{g}_{24i}{∆Y}_{t-i}+{\sum }_{i=0}^j{g}_{25i}{Y}_{t-1}^2\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\sum }_{i=0}^j{g}_{26i}∆{ORP}_{t-i}+{\sum }_{i=0}^j{g}_{27i}∆{ORN}_{t-i}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\sum }_{i=0}^j{g}_{28i}∆{GPRP}_{t-i}+{\sum }_{i=0}^j{g}_{29i}∆{GPRN}_{t-i}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\sum }_{i=0}^j{g}_{30i}∆{(GPR\times OR)P}_{t-i}+{\sum }_{i=0}^j{g}_{31i}∆{(GPR\times OR)N}_{t-i}\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+e_{4t}\end{array}$$

(17)

In Equations (16) and (17), Δ is the first difference operator, and e_3t and e_4t are the white noise error terms. In Equation (16), b₁₀ is the intercept term. j is the optimal lag length (selected by AIC). b_11i–b_19i will be used to test for cointegration (the Bound test) with a null hypothesis of (H₀: b₁₁ = b₁₂ = b₁₃ = b₁₄ = b₁₅ = b₁₆ = b₁₇ = b₁₈ = b₁₉ = 0). Then, b₁₂, b₁₃, b₁₄, b₁₅, b₁₆, b₁₇, b₁₈, and b₁₉ will be normalized by b₁₁ to estimate long run effects. In Equation (17), g₂₀ is the coefficient of the error correction term (ECT_2t−1), and g_23i–g_31i are short-run effects.

4. Data Analysis

Figure 1 presents the annual time-series data in natural logarithms for CO₂ emissions and GDP per capita, GPR, and OR in Saudi Arabia from 1970 to 2024. Over the sample period, CO₂ emissions exhibit an initial trend from approximately 2.68 in 1970 to a peak of 3.46 in 1980, followed by a gradual decline until 1986 and stabilization between 2.68 and 3.32 from 1986 onward. Examining co-movement, GDP per capita shows a steady increase from 12.19 in 1970 to 12.67 in 1974 and then declines until 1989, after which it remains constant. The declining trend of GDP per capita suggests that post-1980 economic growth coincided with a declining CO₂ emissions trend, which indicates a potential decoupling of emissions from GDP growth consistent with the EKC hypothesis. GPR exhibits a minimum value of −4.13 in 1972 and the highest value of −0.14 in 1990. Moreover, most years show higher geopolitical tensions. The co-movement between GPR and CO₂ emissions is generally negative. For instance, during the 1970s, GPR was more negative, and CO₂ emissions were relatively higher. OR shows a highest peak in 1979 with a value of 4.47 and a drop in 2020 with a value of 2.77. Mostly, OR shows a comparatively stable trend around an average of 3.5 after 1980. During the 1970s, OR shows a relatively higher share in GDP, which is consistent with higher CO₂ emissions per capita during 1970, reflecting a general positive co-movement between OR and CO₂ emissions per capita.

Table 1. VIF analysis.

Variables	Yt	GPRt
Yt
GPRt	1.36
ORt	4.65	2.11

reports the Variance Inflation Factor (VIF) results to examine the presence of possible multicollinearity among the explanatory variables, and all VIFs are found to be less than 10, which indicates that multicollinearity is not a severe issue in the hypothesized model. However, a slightly high VIF (4.65) is found between economic growth and oil rents, which is due to the reason that GDP per capita significantly depends on oil rents in Saudi Arabia, but the VIF is still less than 10, and the estimated coefficients can be interpreted reliably.

Table 2. Unit root test.

Variables	MZa	MZt	MSB	MPT
CO2t	−1.7621	−0.7744	0.4239	11.1716
Yt	−10.5381	−2.2055	0.2074	7.7918
Yt2	−10.1467	−2.1569	0.2010	7.8860
GPRt	−12.9365	−2.4935	0.1876	7.6252
ORt	−13.3973	−2.7138	0.1847	6.5757
GPRPt	−7.5339	−1.6763	0.2288	10.1256
GPRNt	−8.9360	−1.8850	0.2279	9.2553
ORPt	−6.695	−1.7436	0.2675	10.3198
ORNt	−11.3847	−2.1579	0.1868	5.9056
GPRt*ORt	−7.5599	−1.8199	0.2396	11.4396
(GPR*OR)Pt	−12.319	−2.4691	0.1895	5.6595
(GPR*OR)Nt	−7.5027	−1.8734	0.2041	8.8085
ΔCO2t	−24.1416 ***	−3.4333 ***	0.1358 ***	3.0274 ***
ΔYt	−18.7656 **	−2.8623 **	0.1547 **	4.9311 **
ΔYt2	−18.1766 **	−2.8823 **	0.1598 **	4.2360 **
ΔGPRt	−25.4644 ***	−3.7190 ***	0.0985 ***	3.0236 ***
ΔORt	−24.1119 ***	−3.5383 ***	0.1151 ***	2.9595 ***
ΔGPRPt	−24.7747 ***	−3.6015 ***	0.1211 ***	2.9899 ***
ΔGPRNt	−21.6399 **	−2.9617 **	0.1451 **	4.1245 **
ΔORPt	−22.2215 **	−3.0708 **	0.1529 **	4.1410 **
ΔORNt	−23.8892 ***	−3.3180 ***	0.1194 ***	2.9735 ***
ΔGPRt*ORt	−24.5786 ***	−3.3739 ***	0.1243 ***	2.8909 ***
Δ(GPR*OR)Pt	−23.2561 ***	−3.3576 ***	0.1550 **	2.9999 ***
Δ(GPR*OR)Nt	−18.0995 **	−2.8365 **	0.1516 **	4.2188 **

Note: ** and *** depict stationarity at 5% and 1%.

shows the unit root analysis, based on Equations (2)–(5). All variables are found to be non-stationary at their levels. However, ΔCO_2t, ΔGPR_t, ΔOR_t, ΔGPRP_t, ΔORN_t, Δ(GPR*OR)P_t, and ΔGPR_t*OR_t are stationary at 1% level of significance, and the rest of the differenced variables are stationary at 5%.

Table 3. Bound test.

Model	F-Values from the Bound Test	Heteroscedasticity	Serial Correlation	Normality	Functional Form
Linear ARDL	6.7538 ***	1.1500 (0.3186)	0.1836 (0.7804)	0.1802 (0.8591)	2.1141 (0.1201)
Nonlinear ARDL	5.1645 ***	1.5946 (0.2056)	1.1075 (0.5525)	0.8308 (0.7101)	1.8669 (0.2553)

Note: *** shows cointegration at 1% level of significance.

shows the results of the bounds test and diagnostic checks, based on Equations (6) and (16). F-values are more than their upper critical bounds at the 1% level of significance in both models. Thus, cointegration is corroborated in both models. The p-values of all diagnostic tests are found to be more than 0.1 in both models, which indicates that both models are free from any of the tested econometric problems. Thus, the residuals have constant variance. Serial correlation is not detected in the residuals. The residuals are normally distributed. The functional forms show appropriate model specifications. Thus, the results demonstrate that both models are well specified.

Figure 2 shows CUSUM and CUSUMsq stability tests for both models, which test the structural stability of a model’s estimated coefficients over the sample period. The estimated lines of both tests remain inside the 5% significance bounds in all sample periods, which corroborate stable estimated parameters over time.

Table 4. Long run results.

Regressors	Linear ARDL	Nonlinear ARDL
Yt	34.5214 *** (0.0264 *)	39.5416 *** (0.0052 *)
Yt2	−1.6219 ** (0.0567 *)	−1.8658 ** (0.0669 *)
ORt	0.1972 *** (0.0125 *)
ORPt		0.2175 ** (0.0654 *)
ORNt		0.0574 ** (0.0947 *)
GPRt	−0.1676 ** (0.0847 *)
GPRPt		−0.2542 *** (0.0165 *)
GPRNt		−0.1004 *** (0.0067 *)
GPRt*ORt	0.1397 ** (0.0954 *)
(GPR*OR)Pt		0.1921 *** (0.0245 *)
(GPR*OR)Nt		0.0676 * (0.2465 *)
Intercept	−138.4380 ** (0.0541 *)	−56.5241 * (0.1547 *)

Note: (p-values). *, **, and *** represent level of significance at 10%, 5%, and 1%.

presents the long-run estimates obtained from the ARDL and NARDL, based on Equations (6) and (16). The EKC is corroborated in both models with the positive parameters of Y_t and the negative parameters of Y_t². Thus, both models explain that carbon emissions initially rise with increasing income and later decline after reaching the income threshold. In the ARDL results, OR_t raises CO_2t, and a 1% increase in OR increases emissions by 0.1972%. In NARDL results, a 1% rise in ORP_t increases emissions by 0.2175%, and a 1% decrease in ORNt decreases emissions by 0.0574%. The positive shocks of OR carry about a four-times greater effect compared to the effect of negative shocks of OR, which indicates the presence of an asymmetric impact of OR. In addition, the Wald test is applied with the null hypothesis of equal effects of both positive and negative shocks of OR (symmetrical effects). The estimated Chi-square is 184.56 with a p-value less than 0.01. Thus, the null hypothesis is rejected, and asymmetric effects are confirmed. In the ARDL results, a 1% rise in GPR reduces emissions by 0.1676%. In NARDL results, a 1% rise in GPRP_t decreases emissions by 0.2542%. On the other hand, a 1% decrease in GPRN_t increases emissions by 0.1004%. Thus, the negative effect of GPRP_t is found to be more than 2.5 times greater than the effect of GPRN_t. Thus, increases in GPR have a stronger environmental impact than decreases in GPR. The asymmetric effects of GPR are also verified with the Wald test and validate asymmetry with the estimated Chi-square = 121.36 and a p-value < 0.01. In the ARDL results, a 1% simultaneous rise in GPR_t*OR_t may amplify emissions by 0.1397%. In NARDL results, a 1% simultaneous rise in (GPR*OR)P_t may increase carbon emissions by 0.1921%. However, the effect of (GPR*OR)N_t is found to be insignificant. The asymmetric effects of GPR_t*OR_t are also tested with the Wald test, which validates asymmetry with the estimated Chi-square = 84.54 and a p-value < 0.01. Thus, a simultaneous increase in oil rents and GPR cannot help reduce carbon emissions.

Table 5. Short run results.

Regressors	Linear ARDL	Nonlinear ARDL
ΔCO2t−1	0.2386 ** (0.0947 *)	0.5947 *** (0.0498 *)
ΔYt	22.9578 *** (0.0254 *)	26.5469 *** (0.0056 *)
ΔYt2	−1.0833 *** (0.0297 *)	−1.2526 *** (0.0349 *)
ΔORt	0.4719 *** (0.0249 *)
ΔORPt		0.4491 ** (0.0647 *)
ΔORNt		0.1704 *** (0.0244 *)
ΔGPRt	−0.4021 *** (0.0439 *)
ΔGPRPt		−0.5123 *** (0.0000 *)
ΔGPRNt		−0.0787 * (0.1269 *)
Δ(GPRt*ORt)	0.1295 *** (0.0009 *)
Δ(GPR*OR)Pt		0.1504 *** (0.0067 *)
Δ(GPR*OR)Nt		0.1138 * (0.8168 *)
ECTt−1	−0.6974 *** (0.0005 *)	−0.7831 *** (0.0000 *)

Note: (p-values). *, **, and *** represent level of significance at 10%, 5%, and 1%.

presents the short-run estimates obtained from the ARDL and NARDL, based on Equations (8) and (17). The negative coefficients of ECT_t−1 show the existence of short-run relationships in both ARDL and NARDL models, with the speed of adjustment for the NARDL model (−0.7831) and the ARDL model (−0.6974). Thus, the ECM works better in the NARDL model. The coefficients show that both models’ short-run deviations toward equilibrium can be settled with a speed of about 70–78% per year. Moreover, the coefficients of lagged CO₂ emissions are positive. Thus, past carbon emissions are partially responsible for current period emissions in both models. Moreover, the EKC is again corroborated by the estimated parameters of ΔY_t and ΔY_t² in both ARDL and NARDL models. ΔOR_t amplifies emissions with a coefficient of 0.4719, which is greater than the long-run coefficient. Thus, increases in OR produce larger instantaneous spikes in emissions before adjusting to the long-run equilibrium. In NARDL results, a 1% increase in ΔORP_t may increase emissions by 0.4491%, and a 1% decrease in ΔORN_t may decrease emissions by 0.1704%. Thus, the spikes of positive shocks of OR are stronger than the spikes of negative shocks of OR, which validates asymmetry. The asymmetric effects of ΔORP_t and ΔORN_t are also tested with the Wald test, and the estimated Chi-square = 112.33 and a p-value < 0.01 validate the asymmetry.

In ARDL results, ΔGPR_t has a negative effect on ΔCO₂t, and a 1% rise in ΔGPR_t might reduce 0.4021% of carbon emissions. In NARDL results, the effect of ΔGPRP_t is found to be negative, but the effect of ΔGPRN_t is insignificant. The asymmetrical effects of ΔGPRP_t and ΔGPRN_t are tested with the Wald test, and the estimated Chi-square = 156.35 and a p-value less than 0.01 validate the asymmetry. Thus, increasing GPR would reduce carbon emissions, but decreasing GPR could not amplify emissions, which validates the short-run asymmetric effect. So, the environmental influence of decreases in GPR may take a longer time to materialize. A 1% rise in GPRP_t may decrease emissions by 0.5123%. In ARDL results, GPR_t*OR_t has a positive coefficient, and a 1% simultaneous rise in GPR_t*OR_t may amplify emissions by 0.1295%. In NARDL results, a 1% simultaneous change in (GPR*OR)P_t may increase carbon emissions by 0.1504%. However, the effect of (GPR*OR)N_t is found to be insignificant, which corroborates short-run asymmetry in this relationship. In addition, the asymmetric effects of Δ(GPR*OR)P_t and Δ(GPR*OR)N_t are also tested with the Wald test, which validates asymmetry with the estimated Chi-square = 96.57 and a p-value < 0.01.

5. Discussions

The results of both ARDL and NARDL corroborate the EKC, and this result is consistent with Alsubaie et al. (2025), who have similarly confirmed the EKC for Saudi Arabia. Moreover, Ma et al. (2022) have also validated the EKC for developing economies. The findings of EKC validity corroborate the efforts of Saudi Vision 2030, which targets renewable energy and sector transformation in the economy. Thus, rising GDP per capita in the Kingdom has no environmental problems. The results of ARDL and NARDL corroborate that the coefficient of OR is less than 1, and the impact of increasing OR is higher than the effect of decreasing OR. This result aligns with Prospect Theory, which explains different responses to gains and losses.

OR is a major proportion of GDP, government revenue, and exports in the Kingdom. Rising oil rents are viewed as gains, which can be invested in oil production, industrial expansion, public spending, and energy-intensive development. Thus, increasing OR can contribute to carbon emissions as all these activities are energy-intensive in Saudi Arabia. However, the effect is found to be less than one, which explains that the response of increasing carbon emissions is less than the rise in OR. Thus, some part of increasing OR is also invested in renewable energy, technologies, and sectors. For instance, the Saudi economy is nowadays significantly spending its income on the tourism sector and renewable energy projects, which can slow down the growth of emissions compared to the growth of OR. Moreover, the impact of decreasing OR is smaller than the impact of rising OR. Falling oil rents are viewed as losses in the Saudi economy. Thus, policymakers may try to protect output, revenues, and economic stability instead of investing in renewable markets. Therefore, carbon emissions will not decline proportionately to OR declines. This finding reflects the dependence of the structure of the Saudi economy on the oil sector, where OR positive shocks quickly translate into greater environmental damage compared to environmental improvement with OR negative shocks. This result also highlights the importance of diversification. The result of the positive relationship between NRR and emissions is aligned with the past literature (Iqbal et al., 2025; Lin et al., 2024; Du & Wang, 2023). Moreover, the asymmetric results are also consistent with the findings of W. Zhao et al. (2021).

The negative association between GPR and emissions is also found to be non-proportional (less than 1), and the impact of increasing GPR is higher than the impact of decreasing GPR. The negative relationship explains that increasing GPR may immediately disrupt trade, transport, consumption, investments, and industrial activities, which can reduce energy consumption and emissions in Saudi Arabia. However, the environmental benefits are smaller than the economic losses because of fossil-fuel dependence. The smaller effect of decreasing GPR compared to the effect of increasing GPR is in line with Prospect Theory, as the effect of gain (decreasing GPR) is smaller than the loss (increasing GPR). The negative effect of GPR on emissions aligns with the literature (Hashmi et al., 2022; Husnain et al., 2022; Ma et al., 2022; Kızılkaya et al., 2024; Villanthenkodath & Pal, 2024; Nawaz et al., 2023; Saadaoui et al., 2024).

The interaction between GPR and oil rents positively affects carbon emissions. Thus, under conditions of both higher OR and rising GPR, policymakers may give stronger priority to oil revenue generation, energy security, and short-term economic stability instead of caring for environmental problems. Moreover, higher OR provides the financial incentive to expand extraction and carbon-intensive activities in the oil sector, which is carbon-intensive. On the other hand, higher GPR will shift the intention of policymakers toward fiscal sustainability, as OR is a major contributor to public spending. Thus, increasing both GPR and OR can contribute to carbon emissions as both can give rise to carbon-intensive activities in the oil sector and may discourage investment in the renewable sector (L. Chen et al., 2023). Nevertheless, the insignificant impact of both ORN_t and GPRN_t on carbon emissions explains that falling both factors simultaneously cannot help reduce carbon emissions, which corroborates the ratchet effects in the case of decreasing both factors. For instance, simultaneously increasing OR and GPR increases carbon emissions, but simultaneously decreasing both factors cannot reverse the environmental condition of the country by reducing carbon emissions.

In a limitation of this study, it is acknowledged that the generalizability of the findings is specific to Saudi Arabia, which is a rentier economy with extreme oil dependence, centralized governance, and high exposure to regional geopolitical conflicts. However, these structural characteristics may not be replicated in other resource-rich or resource-poor economies, which restricts the applicability of the findings to other resource-rich or developing economies. For instance, the asymmetric effects of oil rent shocks and geopolitical risk on emissions could differ in countries with diversified energy portfolios, stronger environmental institutions, and different geopolitical risk profiles. Consequently, the results can be interpreted as country-specific instead of universally applicable. However, future comparative studies across multiple oil-dependent economies would help establish the external validity of the asymmetric relationships.

6. Conclusions

In the Saudi oil-dependent economy, the interaction of economic growth, OR, and GPR may shape environmental outcomes. Thus, the present research investigates the effect of GPR and OR in the EKC framework by using a dataset from 1970 to 2024 in Saudi Arabia. In addition, Prospect Theory suggests investigating the asymmetric nexus between OR and emissions. Thus, NARDL is utilized. The results substantiate the long- and short-run EKC, which corroborates that environmental benefits can be realized after a threshold GDP per capita. OR is identified as a major driver of emissions. In the asymmetric analysis, positive shocks in OR increase CO₂ emissions more strongly compared to the effect of negative shocks. This result corroborates Prospect Theory with asymmetric effects of OR. Thus, increasing OR may contribute to more CO₂ emissions compared to the decline in emissions with decreasing OR, which corroborates a slight ratchet effect of energy consumption behavior with rising income due to rising OR. Moreover, rising GPR reduces carbon emissions, and falling GPR increases emissions in the long run. Nevertheless, falling GPR shows a short-run insignificant effect, but positive shocks in GPR also reduce emissions in the short run. The simultaneous positive shocks in both OR and GPR (the interaction effect) raise emissions, and the simultaneous negative shocks in both OR and GPR do not affect carbon emissions. Thus, GPR significantly and adversely moderates environmental degradation with rising OR. However, GPR could not moderate the positive environmental outcomes with declining OR. Overall, the effects of OR, GPR, and their interaction are asymmetric. The oil dependence of the Saudi economy has resulted in environmental problems. Moreover, the combined effect of a simultaneous increase in OR and GPR also shows environmental concerns. However, GPR shows a negative direct effect on emissions. Thus, GPR is directly responsible for the slowdown of economic activities, which helps improve environmental quality at the cost of economic growth.

6.1. Policy Implications

The results show that positive shocks in OR increase emissions more than the decline in emissions associated with negative shocks in OR, which emphasizes the existence of emission-intensive activities even during periods of declining OR. Thus, policymakers should focus on proactive emission control measures. For instance, the Ministry of Energy and the Saudi oil company (Aramco) should enforce stricter emission intensity standards for upstream oil extraction and use clean technologies across all oil production facilities. Additionally, the government should invest in green projects, including solar and wind technologies, green hydrogen production, and non-oil sectors. Rising GPR helps reduce carbon emissions due to slower economic activity, but declining GPR again increases emissions (a rebound effect). During high-GPR periods, to further enhance environmental benefits, the government should invest in renewable infrastructure to develop renewable capacity, which could also help reduce emissions during low-GPR periods. During low-GPR periods, the government should use countercyclical green transition policies. For instance, the government should increase carbon efficiency standards by introducing dynamic carbon pricing mechanisms and pollution taxes and strengthen environmental monitoring. This action may provide more fiscal space to invest in renewable energy projects and expand green infrastructure. Moreover, the government can also provide fiscal incentives to the private sector for adopting renewable technologies to offset the pollution-intensive effects of economic growth. Moreover, a simultaneous rise in OR and GPR increases emissions, which shows that high OR during high GPR has been largely directed toward fossil-fuel-based activities instead of renewable energy projects. Thus, the government should adopt integrated energy, environmental, and risk-management policies to reduce the pollution-enhancing effects of oil rents during high-GPR periods. For instance, during high GPR, private investments in renewable projects may decline. However, the government has a buffer fiscal space due to high OR, which can be invested in renewable energy development, green public industrialization projects, and sustainable infrastructure projects.

6.2. Limitations and Future Directions

The study faces limitations due to the limited availability of GPR data for other GCC countries, which restricts the applicability of the findings to other GCC economies. Future research may construct self-generated or country-specific GPR indices for other oil-rich economies to conduct comparative studies, which could improve the external validity and generalization power of the findings. Moreover, the study relies on aggregate national-level data, which reduces the ability to capture sector-specific environmental responses to fluctuations in GPR and oil rents. For instance, the oil sector, transportation, manufacturing, non-oil industries, electricity generation, and service sector may respond differently to changes in GPR and oil-related revenues. Future studies could conduct sector-specific analyses to provide a more detailed understanding of the heterogeneous environmental responses to fluctuations in GPR and oil rents across different sectors, which would offer more targeted policy implications for designing sector-specific environmental and energy strategies.