Do Natural Resource Rents Hinder or Promote Renewable Energy Investment? The Moderating Role of Institutional Quality and Financial Development in Resource-Rich Developing Economies

The discourse on the natural resource curse originated with the seminal cross-country analysis by Sachs and Warner [3], demonstrating that an abundance of natural resources, particularly when measured by primary exports as a percentage of Gross Domestic Product (GDP), negatively correlates with per capita economic growth. Subsequent research has contested the universality of this finding [15], yet the preponderance of evidence, particularly at the subnational level and over prolonged durations, corroborates the assertion that resource revenues precipitate Dutch disease effects, undermining tradable manufacturing and knowledge-intensive sectors [16, 17]. Torvik [18] and Robinson et al. [19] suggested the mechanisms by which resource rents lead to institutional deterioration, encompassing rent-seeking, reduced state capacity, and weakened accountability. Mehlum et al. [20] empirically demonstrated that the growth impacts of resource wealth are contingent upon institutional quality.

The concept of the resource curse has been utilized in energy economics literature to explain why fossil fuel exporters consistently fall behind in the adoption of renewable energy. The Dutch disease mechanism directly impacts the energy sector: abundant fossil fuel revenues reduce the relative price of energy, weaken the incentive for efficiency improvements, and generate fiscal revenues that supplant extensive economic transformation. Fossil fuel rents create powerful stakeholders, including state energy companies, fossil fuel laborers, and dependent financial authorities, who strongly resist energy transition initiatives [21, 22]. Gorji and Martek recently analyzed the impact of renewable energy policies on technology implementation in 20 developed and 20 developing oil-producing countries, revealing that the resource curse considerably diminishes the efficacy of policy support for renewables in resource-dependent nations [23]. Destek et al. [24] examined the influence of economic complexity in mitigating the resource curse across a comprehensive panel, discovering that diversified economic structures diminish the adverse relationship between resource dependency and sustainable development outcomes. Ragmoun and Ben-Salha [25] further demonstrated that oil rents adversely affect environmental quality in Saudi Arabia, whereas renewable energy consumption and green technological innovation improve it, providing country-level evidence that resource wealth impedes clean energy transitions in the absence of technological upgrading. Complementing this sectoral evidence, Sobirov et al. [26] tested the Environmental Kuznets Curve in ten Association of Southeast Asian Nations (ASEAN) economies using FMOLS, DOLS, and CCR estimators, showing that scale effects from GDP expansion and trade openness dominate composition effects, and that Foreign direct investment (FDI) exerts no statistically significant aggregate effect on CO₂ emissions. These findings reinforce the view that structural dependence on fossil fuels and regulatory gaps remain core barriers to sustainable growth in resource-dependent developing regions.

The empirical investigation into the factors influencing renewable energy investment includes microeconomic, sectoral, and macroeconomic analysis. Sadorsky [6] established a cointegrating relationship between per capita income and renewable energy consumption at the national level, employing panel data from emerging economies. Apergis and Payne [7] expanded the analysis utilizing FMOLS and DOLS, confirming the positive income elasticity of renewable energy demand. Shahbaz et al. [27] identified financial development as a catalyst, arguing that more sophisticated capital markets diminish the financing costs associated with capital-intensive renewable projects. Asongu and Odhiambo [10] employed Generalized Method of Moments (GMM) and quantile regression on African panel data to confirm that financial accessibility and institutional governance are key determinants of renewable energy uptake. Prempeh et al. [28] validated the environmental Kuznets curve hypothesis and renewable energy within Sub-Saharan Africa, demonstrating through Dumitrescu and Hurlin causality tests that financial development and governance jointly determine renewable energy consumption within a panel of 38 Sub-Saharan Africa (SSA) countries. Marques et al. [8] conducted a cross-national regression analysis to identify the determinants of renewable electricity, emphasizing energy dependency, carbon emissions commitments, and market structure as key factors. Polzin et al. [9] implemented a more nuanced investment-flow perspective and show that institutional investors whose participation depends on regulatory transparency and governance standards account for an increasing share of global renewable energy financing. Egli et al. [29] established that country risk, closely associated with institutional quality, substantially elevates the cost of capital for renewable projects, therefore impeding investment.

Recent empirical research has further corroborated these links. Anton and Nucu [30] utilized panel data from European Union (EU) nations and illustrated that financial development, quantified by domestic credit and stock market capitalization, substantially enhances renewable energy consumption, even when accounting for income and energy prices. Li et al. [31] examined productive capacity, renewable energy investments, and climate mitigation technologies alongside fiscal policy challenges in the top-10 resource-exporting economies, finding that responsible resource production and consumption require coordinated fiscal frameworks to support sustainable energy transitions. Imtiaz et al. [32] investigated the impact of debt, renewable energy, and energy imports on Pakistan’s carbon emissions, demonstrating that leapfrogging toward sustainability is feasible when renewable energy adoption is coupled with prudent debt management and reduced fossil fuel import dependency. Haldar and Sethi [33] presented cross-national evidence indicating that institutional quality substantially decreases CO₂ emissions by fostering renewable energy utilization in developing nations, employing CCEMG and AMG estimators that address CD. In a methodologically parallel study, Shanyazov et al. [34] applied second-generation Cross-sectionally Augmented Autoregressive Distributed Lag (CS-ARDL) and AMG estimators to G7 economies over 2000 to 2022, showing that ICT goods exports and R&D expenditure both contribute to lowering energy intensity, with long-run elasticities of −0.082 and −0.155 respectively, offering evidence that supports this study’s findings on the role of innovation-enabling institutions in the energy transition. Rafiq et al. [35] validated the beneficial impact of institutional quality on the advancement of renewable energy in Organisation for Economic Co-operation and Development (OECD) countries, utilizing rigorous panel unit root and cointegration methodologies. In a complementary comparative Bayesian analysis of 22 developing and 26 developed economies, Nguyen and Le [36] demonstrated that both economic freedom and institutional quality positively drive sustainable development outcomes, with institutional quality exerting a particularly strong effect in developing economies, reinforcing the centrality of governance for development and environment objectives.

Tambari et al. [37] studied the connection between oil prices and renewable energy investment in African net oil-importing and net oil-exporting nations, revealing that oil price volatility exerts asymmetric effects on renewable energy investment based on a country’s net trade status. Mahmood et al. [38] investigated institutional quality, renewable energy investment, financial development, and ecological risks in ASEAN nations utilizing AMG and CCEMG estimators, demonstrating that institutional quality and financial development markedly enhance environmental outcomes, whereas renewable energy investments exert a more pronounced effect on reducing per capita carbon emissions over the long term. This literature together indicates that institutional quality and financial development are both theoretically coherent and empirically substantiated factors affecting renewable energy investment, but extensive panel evidence from resource-rich developing nations remains limited, and quantile-based evidence on how these effects vary across the conditional distribution of renewable energy investment is virtually absent.

The theoretical framework incorporates three elements of economic theory: the resource curse hypothesis, the political economy of energy transition, and financial intermediation theory. The following hypotheses are hereby proposed:

Hypothesis 1 (H1): Natural resource rents have a negative and statistically significant impact on renewable energy investment in resource-abundant developing nations.

Hypothesis 2 (H2): Institutional quality positively moderates the relationship between natural resource rents and renewable energy investment, mitigating the crowding-out effect.

Hypothesis 3 (H3): Financial development has a positive and statistically significant impact on renewable energy investment, mediated by cost-of-capital and risk-pooling mechanisms.

The econometric framework addresses four key issues in sequence: (i) non-stationarity in panel time-series variables; (ii) the presence of a long-run cointegrating relationship; (iii) efficient estimation of long-run elasticities and short-run dynamics under slope heterogeneity and CD; and (iv) distributional heterogeneity in long-run relationships across the conditional distribution of renewable energy investment. The analysis is conducted as follows.

Before performing unit root and cointegration tests, it is crucial to determine whether the panel data display CD, as this greatly affects the choice of subsequent tests and estimators. The Pesaran [45] CD test is employed; it is suitable for both large-N and large-T panels and has superior power characteristics relative to the Breusch and Pagan LM test [46]. The rejection of the null hypothesis of no CD necessitates the use of second-generation panel unit root and cointegration tests.

$C D=\sqrt{\frac{2 T}{N(N-1)}} \sum_{i=1}^{N-1} \sum_{j=i+1}^N \rho_{i j} \sim N(0,1)$              (1)

In light of the evidence of CD, the Cross-sectionally Augmented Im, Pesaran, and Shin (IPS) (CIPS) test proposed by Pesaran [47] is applied due to its robustness to both CD and slope heterogeneity. This test extends the standard ADF regression by including cross-sectional averages of lagged levels and first differences. The CIPS statistic is computed as the simple average of the individual cross-sectionally augmented Dickey-Fuller (CADF) statistics. For comparison purposes, the first-generation IPS test [48] is also reported, despite its limitations under CD.

$\Delta y_{i t}=\alpha_i+\beta_i y_{i, t-1}+\gamma_i \overline{y_{t-1}}+\sum_{j=0}^p \delta_{i j} \Delta \overline{y_{t-j}}+e_{i t}$         (2)

The Westerlund [49] panel cointegration test is employed to investigate long-run cointegrating relationships by testing the null hypothesis of no cointegration within an error correction model framework. The test produces four statistics: Gτ and Gα (group-mean statistics) and Pτ and Pα (panel statistics). It is preferred over the Pedroni [50] and Kao [51] tests because it allows for serially correlated errors, heterogeneous short-run dynamics, and CD when bootstrap critical values are applied [52]. Both asymptotic and bootstrap p-values based on 800 replications are reported.

$\begin{gathered}\Delta y_{i t}=\delta_i^{\prime} d_t+\alpha_i\left(y_{i, t-1}-\beta_i^{\prime} x_{i, t-1}\right)+\sum_{j=1}^p \alpha_{i j} \Delta y_{i, t-j} +\sum_{j=0}^q \gamma_{i j} \Delta x_{i, t-j}+\varepsilon_{i t}\end{gathered}$              (3)

In light of the cointegration evidence, long-run coefficients are estimated utilizing two complementary estimators: The CCEMG estimator proposed by Pesaran [11] and AMG estimator developed by Eberhardt and Bond [12] are employed. Both estimators accommodate slope heterogeneity, allowing coefficients to vary across countries, as well as CD. The CCEMG approach augments the regression by including cross-sectional averages of the dependent variable and all regressors to capture unobserved common factors, with the mean group estimate obtained as the simple average of country-specific coefficients. The AMG estimator operates in two stages: it initially estimates a common dynamic process using a first-differenced pooled regression augmented with year dummies, then subsequently subtracts this process prior to estimating individual country regressions. Both estimators exhibit consistency as N and T tend towards infinity and yield asymptotically normal mean group estimates.

The empirical model is specified as:

$\begin{aligned} \ln \left(R E I_{i t}\right)=\alpha_{0 i} & +\alpha_{1 i} N R R_{i t}+\alpha_{2 i} I Q_{i t}+\alpha_{3 i} F D_{i t} \\ & +\alpha_{4 i}\left(N R R_{i t} \times I Q_{i t}\right) \\ & +\alpha_{5 i} \ln \left(G D P P C_{i t}\right) \\ & +\alpha_{6 i} T R A D E_{i t}+\alpha_{7 i} \operatorname{CO2PC} C_{i t} \\ & +\alpha_{8 i} F D I_{i t}+v_{i t}\end{aligned}$            (4)

For estimation, all variables enter in their natural levels except for renewable energy investment and GDP per capita, which are log-transformed to linearise multiplicative effects, and natural resource rents, which are rescaled from the percentage form (0-100) reported in the World Development Indicators to a 0-1 share of GDP. This rescaling ensures that the estimated semi-elasticity ∂ ln(REI)/∂ NRR is interpretable as the proportional change in renewable investment associated with a unit (i.e. 100-percentage-point) change in resource dependence, so that a 10-percentage-point increase in NRR translates into a 0.1-unit shift in the rescaled regressor. The interaction term NRR × IQ is constructed after this rescaling, so that the implied governance threshold IQ* = −β₁/β₄ is dimensionally consistent and economically interpretable. Without this rescaling, a one-percentage-point increase in resource rents would translate into an implausibly large response in renewable investment; the share-based scaling adopted here is consistent with the standard practice in the resource-curse literature.

To ensure robustness, the group-mean FMOLS estimator [53, 54] and the group-mean DOLS estimator [55] are also presented. FMOLS addresses serial correlation and endogeneity in the regressors using non-parametric kernel estimation, while DOLS mitigates endogeneity by incorporating leads and lags of the first differences of the regressors into the regression. These estimators exhibit consistency in the context of cointegration and serve as a valuable benchmark for comparison with the CCEMG and AMG estimates. Empirical applications of these cointegration techniques to the energy-economic growth nexus in developing economies are exemplified by Hdom and Fuinhas [56] and Ouedraogo [57].

Conventional mean-based panel estimators such as CCEMG, AMG, FMOLS, and DOLS provide a single average elasticity for the entire panel and are therefore silent on whether the determinants of renewable energy investment exert different effects on countries situated at different points of the conditional distribution. To address this limitation, this study implements the MMQR with fixed effects developed by Machado and Silva [13]. Unlike traditional quantile regression, the MMQR estimator allows individual fixed effects to influence the entire conditional distribution rather than acting only as location shifters, which is essential when unobserved heterogeneity is correlated with both the level and the dispersion of the dependent variable, as is typical in macro panels of resource-rich developing economies.

Formally, the conditional location-scale model underlying the MMQR is given by:

$Y_{i t}=\alpha_i+X_{i t}^{\prime} \beta+\left(\delta_i+Z_{i t}^{\prime} \gamma\right) U_{i t}$            (5)

where $Y_{\text {it}}$ denotes the natural logarithm of renewable energy investment, $X_{\mathrm{it}}$ is the vector of regressors specified in Eq. (4), $Z_{\text {it}}$ is a vector of known differentiable transformations of the components of $X_{\mathrm{it}}, \alpha_{\mathrm{i}}$ and $\delta_{\mathrm{i}}$ capture the individual location and scale fixed effects, and $U_{\mathrm{it}}$ is an i.i.d. disturbance term that is statistically independent of $X_{\mathrm{it}}$. Under the standard regularity conditions of Machado and Silva [13], the conditional quantile function of Yᵢₜ is recovered as:

$Q_Y\left(\tau \mid X_{i t}\right)=\left(\alpha_i+\delta_i q(\tau)\right)+X_{i t}^{\prime} \beta+Z_{i t}^{\prime} \gamma q(\tau)$         (6)

where $q(\tau)$ is the $\tau-$ th quantile of the standardized error $U_{\mathrm{it}}$. The coefficient on each regressor at quantile $\tau$ is therefore $\beta+\gamma q(\tau)$, so that $\gamma$ measures the heterogeneous component of the effect across the conditional distribution while $\beta$ captures the average location effect. Eq. (6) is estimated at the 10th, 25th, 50th, 75th, and 90th quantiles, which provides a comprehensive characterization of how the effects of natural resource rents, institutional quality, financial development, and the NRR × IQ interaction varies across countries with low, intermediate, and high levels of renewable energy investment. This approach is particularly informative in the present setting because it reveals whether the resource curse and the institutional mediation channel intensify or attenuate as countries advance along the renewable energy investment frontier.

To investigate directional causal linkages among the variables, this study employs the Dumitrescu and Hurlin [14] panel Granger non-causality test, which is specifically designed for heterogeneous panels and remains valid in the presence of slope heterogeneity and CD, two features that invalidate first-generation panel causality tests in macro datasets such as ours. Unlike the homogeneous Granger causality test that imposes the restrictive assumption that the causal relationship is identical across all cross-sectional units, the Dumitrescu and Hurlin procedure allows the autoregressive parameters and the regression coefficients to differ across countries while testing the joint null hypothesis of no Granger causality from variable X to variable Y for any country in the panel against the alternative that causality runs from X to Y in at least one country.

The test is based on country-specific Granger causality regressions of the form:

$y_{i t}=\alpha_i+\sum_{k=1}^K \gamma_{i k} y_{i, t-k}+\sum_{k=1}^K \beta_{i k} x_{i, t-k}+\varepsilon_{i t}$         (7)

from which the individual Wald statistics $W_{\mathrm{i}, \mathrm{t}}$ for the joint nullity of all $\beta_{\mathrm{ik}}$ are computed. The Dumitrescu and Hurlin average statistic is then defined as:

$\bar{W}=\frac{1}{N} \sum_{i=1}^N W_i$          (8)

and its standardized counterpart $\tilde{Z}$ converges to a standard normal distribution under the null hypothesis as both T and N tend towards infinity, while a semi-asymptotic version $\tilde{Z}$ is also available for finite T. A rejection of the null indicates that there is a Granger causal relationship from X to Y for a non-trivial fraction of countries in the panel, which constitutes the empirically relevant interpretation in heterogeneous macro samples. Causality is tested in both directions for all theoretically relevant variable pairs, namely NRR and REI, IQ and REI, FD and REI, NRR and CO2PC, and GDPPC and REI, in order to discriminate among unidirectional, bidirectional, and feedback structures and thereby illuminate the temporal precedence underlying the long-run elasticities estimated by the CCEMG, AMG, FMOLS, DOLS, and MMQR estimators.

The empirical findings of this study carry several dimensions of policy significance for resource-rich developing countries navigating the global energy transition. This section contextualizes the findings within the current empirical literature and formulates specific policy recommendations.

The negative and statistically significant coefficient for NRR (−0.412, CCEMG) in the resource curse channel (H1), corroborated by the monotonically increasing magnitude of the NRR coefficient across MMQR quantiles, provides robust panel evidence that natural resource rents directly impede renewable energy investment in developing nations and that this drag intensifies as countries advance toward the renewable investment frontier. This finding is consistent with the foundational resource curse literature of Sachs and Warner [3] and the theoretical models proposed by Torvik [18] and Robinson et al. [19], illustrating how resource dependence fosters rent-seeking, diminishes state capacity, and undermines accountability. The findings of this study extend this reasoning to the domain of energy transition. The magnitude of the NRR coefficient broadly corresponds with the findings of Gorji and Martek [23], who reported that the resource curse substantially undermines the efficacy of renewable energy policy in oil-producing developing nations, based on panel data from 2010 to 2020. Likewise, Destek et al. [24] observed that resource dependence obstructs sustainable development outcomes unless counterbalanced by economic complexity, which is consistent with the finding that the crowding-out effect can be mitigated by institutional quality. Similarly, Ragmoun and Ben-Salha [25] provide country-level evidence from Saudi Arabia that oil rents degrade environmental quality, while renewable energy and green technological innovation improve it, reinforcing the mechanism by which fossil fuel dependence impedes clean energy adoption. The unidirectional Granger causality from NRR to REI documented by the Dumitrescu and Hurlin test substantiates the hypothesized causal direction posited in the Dutch disease literature [16, 17], confirming that this relationship is not driven by reverse feedback but rather reflects a temporally precedent causal pressure exerted by the fossil fuel sector on the renewable sector. This creates a self-reinforcing cycle: fossil fuel revenues finance public services, reducing the motivation for economic diversification; they also subsidize energy consumption, distorting market price signals that would typically encourage renewables; and they establish political economy incumbencies that resist clean energy policies [21, 22]. To break this cycle, deliberate institutional design is required, rather than merely economic growth or technology diffusion.

The identification of institutional quality as the foremost positive determinant of renewable energy investment (H2), with a CCEMG coefficient of 0.587 and an MMQR coefficient that rises from 0.412 at the 10th quantile to 0.731 at the 90th, signifies the most substantial marginal effect in the model, exceeding that of financial development by nearly 76%. This finding directly supports the foundational work of Mehlum et al. [20], who demonstrated that the growth consequences of resource wealth depend on institutional quality. Li et al. [31], examining the top-10 resource-exporting economies, demonstrated that renewable energy investments and climate mitigation technologies require coordinated fiscal policy frameworks and productive capacity enhancements for responsible resource management, reinforcing the governance and investment nexus identified in the results. Imtiaz et al. [32] provided complementary evidence from Pakistan, showing that renewable energy adoption combined with prudent debt and energy import management facilitates leapfrogging toward sustainability, a finding consistent with the institutional mediation mechanism identified in this study. Haldar and Sethi [33], utilizing the same CCEMG and AMG estimators as this study, confirmed that institutional quality enhances renewable energy consumption and reduces CO₂ emissions in developing economies; this study extends that work by incorporating the interaction between institutional quality and resource rents and by characterizing the distributional heterogeneity of the moderating mechanism through MMQR. Rafiq et al. [35] further corroborated the positive association between institutional quality and renewable energy for OECD nations, indicating that this relationship holds across both developed and developing contexts, while the comparative Bayesian evidence of Nguyen and Le [36] reinforces that institutional quality is particularly decisive for sustainable development in developing economies.

The interaction result indicates that improvements in governance can mitigate the resource curse within the energy sector, and the MMQR estimates further reveal that this mediating effect is strongest where the resource curse itself is strongest, namely at the upper quantiles of the renewable energy investment distribution. The threshold analysis identifies a governance score of IQ* ≈ 1.466 (approximately 2.42 standard deviations above the sample mean) as the tipping point at which resource rents transition from impeding to facilitating renewable investment. The bidirectional Granger causality between IQ and REI documented in Table 9 shows that progress toward this threshold is self-reinforcing, since improvements in governance trigger renewable investment, which in turn generates new constituencies and regulatory capacity that further upgrade institutional quality. This result provides an empirical basis for the design of institutional conditionality in international climate finance instruments, such as the IMF and World Bank Climate Policy Assessment Framework and the EU Global Gateway initiative. Policymakers in countries approaching this governance threshold should consider sovereign green bonds supported by resource revenue allocation as a mechanism to operationalize the institutional mediation channel.

The substantial and statistically significant effect of financial development (H3), reflected in a CCEMG coefficient of 0.334 and a mildly increasing MMQR profile across quantiles, underscores the critical yet underappreciated role of domestic capital market development in the renewable energy transition of developing countries. This finding supports Shahbaz et al. [27], who identify financial development as a significant determinant of energy consumption in China via multivariate framework analysis, and Asongu and Odhiambo [10], who confirm the positive association between financial access and renewable energy consumption in Sub-Saharan Africa utilizing GMM. Prempeh et al. [28] extend this evidence to a broader SSA panel of 38 nations, confirming through the Renewable Energy Environmental Kuznets Curve (REKC) framework that financial development substantially influences renewable energy consumption, with bidirectional causality between financial development and renewable energy further supporting the virtuous cycle identified in the results. Notably, the FD coefficient is smaller than the IQ coefficient, consistent with the findings of Anton and Nucu [30], who identify a substantial although moderate effect of domestic credit on renewable energy consumption in EU nations. The results are further supported by Egli et al. [29], who demonstrate that country risk, intrinsically linked to institutional and financial development, substantially elevates the cost of financing for renewable projects. Mahmood et al. [38], employing AMG and CCEMG estimators on ASEAN data, indicate that financial development substantially affects environmental outcomes via renewable energy channels, although that investigation prioritizes ecological risks over investment itself. Evidence from Shanyazov et al. [34] on G7 economies further confirms, via CS-ARDL estimation, that innovation-related channels such as R&D expenditure and ICT exports are significant long-run drivers of energy-sector outcomes, suggesting that financial deepening operates in tandem with technology and knowledge investments. The relatively lower coefficient of FD compared to IQ supports the notion that financial development facilitates the implementation of investment decisions primarily shaped by political economy and governance factors. Most international climate finance discussions emphasize external financial inflows; the findings suggest that strengthening domestic financial systems through banking sector reform, expanding the green bond market, and de-risking pension funds could yield comparable investment multipliers at lower fiscal cost.

The bidirectional causality between IQ and REI documented in the Dumitrescu and Hurlin test represents a novel finding that enriches the existing literature. While prior research, such as Polzin et al. [9] and Egli et al. [29], highlights the unidirectional effect of governance on investment, the results indicate that renewable energy investment also Granger-causes improvements in institutional quality, which is consistent with the political economy argument that renewable energy transitions generate new constituencies advocating for regulatory reform and transparent governance [22]. The bidirectional FD and REI causality suggests a virtuous cycle, consistent with the financial deepening hypothesis proposed by King and Levine [58]. The unidirectional causality from NRR to CO₂ emissions underscores the environmental costs of resource dependence, corroborating the findings of Tambari et al. [37], which illustrate that oil price fluctuations exert asymmetric effects on renewable energy investment depending on a nation’s net trade position in fossil fuels. Taken together with the sectoral ASEAN evidence of Sobirov et al. [26], these results suggest that the resource, environment, and investment nexus in developing economies is fundamentally shaped by the interaction of structural composition effects, governance quality, and financial depth.