The Impact of Renewable Energy Use, Financial Development, and Industrialization on CO₂ Emissions in Middle-Income Economies—A GMM-PVAR Analysis

Abstract

Middle-income economies contribute significantly to global CO₂ emissions as they pursue economic development, creating an urgent need to understand emission drivers. This article investigates the impact of renewable energy use, financial development, and industrialization on CO₂ emissions in 71 middle-income countries (32 upper-middle income, 39 lower-middle income) between 2002 and 2020. We used the advanced Generalized Method of Moments Panel Vector Autoregression (GMM-PVAR) approach to address endogeneity and reveal complex relationships among the variables. Our findings revealed that renewable energy utilization had no substantial influence on emissions reduction in either upper- or lower-middle-income countries, challenging conventional policy assumptions. Financial development consistently reduces emissions across both income groups (−0.08% and −0.06%, respectively). Industrialization has heterogeneous effects, increasing emissions by 2.03 percent in upper-middle-income countries and with no effect in lower-middle-income countries. Granger causality tests illustrated a bidirectional relationship connecting CO₂ emissions and financial development, whereas no causal link was found between CO₂ emissions and renewable energy use. These findings prove the importance of coordinated policies that strengthen financial systems and sustainable industrial practices.

1. Introduction

In the face of rising global environmental challenges, characterized by unprecedented warming and an increased occurrence of severe meteorological phenomena, proactive measures are necessary to address climate change [1,2]. The Intergovernmental Panel on Climate Change (IPCC) has categorically singled out human activities, particularly fossil-fuel combustion, to be the predominant factor behind global warming, with the consequences becoming increasingly severe on 3.3 to 3.6 billion vulnerable individuals across the globe [3]. The 28th and 29th Conferences of the Parties (COP28 in 2023 and COP29 in 2024) have been crucial to international climate efforts. At COP28, a primary target was to triple global renewable energy capacity and achieve corresponding gains in energy efficiency by 2030, underscoring the importance of expediting the transition toward low-carbon energy systems [4]. Building on this momentum, COP29 advanced the agenda of a just transition by committing to triple climate finance—reaching USD 1.3 trillion annually—for developing countries [5]. Global climate imperatives highlighted the need to examine the complex relationship between renewable energy consumption, financial development, industrialization, and CO₂ emissions, particularly in middle-income economies (MICs), which are pivotal in shaping the trajectory of global climate and economic outcomes.

According to the World Bank, middle-income economies are defined as countries with a Gross National Income (GNI) per capita that falls within a designated range between $1136 and $4465, encompassing major emerging economies such as China, India, Brazil, and Indonesia [6]. These countries are pivotal in the global climate system, driven by their rapid economic growth and rising CO₂ emissions. According to the World Bank data (Figure 1), the share of global CO₂ emissions by the MICs rose by approximately 50 to 60% from 2002 to 2020, reflecting an increase in the growth of economic and Industrial activities. Meanwhile, their share of global gross domestic product (GDP) doubled to 40% in 2022, compared to 20% in 2002, as shown in Figure 2. As economic powerhouses and major emitters, MICs face the challenge of balancing growth and environmental sustainability.

The “middle-income trap” conception highlights the unique challenges faced by MICs. This phenomenon describes economies that stagnate at middle-income levels due to structural barriers. This phenomenon refers to the stagnation of economies at middle-income levels, often due to structural barriers such as insufficient innovation capacity, or over-reliance on resource- and energy-intensive industries [7]. Overcoming this trap necessitates policy responses that prioritize sustainable development. In this context, examining the roles of renewable energy adoption, financial deepening, and industrial development becomes essential for advancing both economic resilience and environmental sustainability [8,9]. The relative impact of MICs on the environment is also hugely relevant, going up by 25 percent, from 40 percent contribution to global CO₂ emissions in 2002 to 65 percent in 2020, as shown in Figure 1. Global energy-related CO₂ emissions increased to a record of 37.8 gigatons in 2024, and MICs, particularly China (12.667 gigatons in 2022) and India (2.955 gigatons in 2023), contributed significantly due to fossil fuel reliance in their industrial and energy sectors [10]. This pattern highlights the importance of reducing emissions in MICs to support the fulfillment of international climate objectives, especially those outlined in the Paris Agreement, which seeks to reduce global warming. While the rapid economic growth observed in MICs has driven significant development gains, it has also been accompanied by rising environmental degradation. This trade-off highlights the urgent need for strategies that decouple economic development from carbon-intensive activities, enabling a more sustainable growth trajectory.

Renewable energy use is a promising direction toward reducing CO₂ emissions and meeting the growing energy needs of numerous MICs [11]. Global renewable electricity generation is estimated to reach over 17,000 TWh by 2030, which represents nearly a 90 percent increase over the value of 2023, most of which is likely to be supplied by solar and wind energy [12]. There is considerable improvement in renewable energy capacity in middle-income countries. By the end of December 2023, India had commissioned 180.79 gigawatts of non-fossil energy plants, including 73.31 GW of solar and 44.73 GW of wind power [13]. Brazil is the most advanced in the use of biofuels, while China is the leading producer of wind energy, contributing to 32 percent of global renewable electricity in 2023 [14,15]. Nevertheless, the barriers to wider integration are caused by high upfront costs, lack of proper grid infrastructure, and policy uncertainties [11,16].

Financial development is essential for enabling renewable energy transitions in middle-income countries (MICs) by supplying the capital and financing mechanisms required for large-scale project implementation [17]. Debt and equity financing rely significantly on efficient financial markets, particularly for technologies where major up-front expenditures and long payback periods are involved. However, several financial systems in MICs face limitations, such as higher borrowing rates and volatile policies, both of which can limit investment in renewable energy projects [18,19]. Donou-Adonsou et al. [20] suggested that financial development plays a significant role in promoting the deployment of renewable energy among the high-income economies, whereas its impact is comparatively limited in low- and middle-income countries. This disparity underscores the need for tailored financial policies. In MICs, growth is often driven through industrialization, a process typically associated with higher volumes of CO₂ emissions. In China, coal-intensive industrial activities have been a primary contributor to the country’s elevated emission levels. Similarly, India, which is undergoing rapid industrialization, recorded a 5.3% increase in emissions in 2024, largely driven by rising electricity demand needed to sustain economic growth [10]. Green innovation and the planned incorporation of renewable energy sources and green technologies have thus become important stresses in eradicating emissions [21,22]. Carbon pricing, renewable energy mandates, and subsidies on energy-efficient technologies are policy tools that can facilitate the energy transition in MICs, enabling MICs to avoid the carbon-intensive development paths of high-income countries. The Brazilian focus on hydropower and biofuels exemplifies a plausible paradigm of sustainable industrial growth [15].

To analytically evaluate such dynamics, it is necessary to have a methodological framework that can handle any possible endogeneity and can capture both the short and long-run dynamics. GMM of Panel Vector Autoregression (PVAR) models provides a suitable empirical framework for examining the interdependent relationships among renewable energy use, financial development, industrialization, and CO₂ emissions in MICs. This approach is particularly well-suited for capturing complex feedback loops and addressing endogeneity issues inherent in dynamic panel data analysis [23,24]. Financial development can drive renewable energy deployment, aiding in emissions reduction. However, the industrialization process may concurrently exacerbate emissions unless it is cautiously integrated into sustainable development initiatives. By using the GMM-PVAR approach, researchers will be able to single out the most important determinants of emissions, determine the impact of policy initiatives, and provide policymakers with empirically supported recommendations in an attempt to reconcile economic growth and sustainability goals.

This study has three main contributions to the literature on the environmental and economic dynamics in the middle-income economies. First, it uses the GMM-PVAR estimation technique to overcome endogeneity issues and bidirectional relations between renewable energy consumption, financial development, industrialization, and CO₂ emissions, addressing the constraints of conventional econometric models. Second, it generates an in-depth discussion and evaluation of the synergistic effect of these factors, which is lacking in the scholarly literature that mostly evaluates them independently. Third, the disaggregation of economies classified as middle-income economies into upper and lower-middle-income brackets allows the study to reveal heterogeneous mechanisms, including industrial carbon lock-in and green finance channels, providing policy recommendations to achieve sustainable development.

This paper is organized as follows: Section 2 presents a literature review, summarizing theoretical and empirical arguments on renewable energy, financial development, industrialization, and CO₂ emissions. In Section 3, the data sources, methodology, and econometric model are discussed. Section 4 presents the results in detail, compares the findings of upper-middle-income countries to lower-middle-income countries. Conclusions and discussions about their implications are outlined in Section 5. Section 6 makes policy suggestions towards balancing economic and environmental sustainability. Section 7 addresses limitations and future research directions.

2. Literature Review

2.1. Renewable Energy and Carbon Emissions

Growing economies are the primary priority for most nations, particularly developing and emerging economies, yet economic expansion is often seen as a substantial contributor to carbon dioxide emissions [25]. At the same time, countries are increasingly aiming to achieve sustainable economic development while safeguarding environmental integrity. Previous research has highlighted the significant role of renewable energy in mitigating carbon emissions [26]. The literature on renewable energy use and carbon emissions reveals negative, neutral, and context-specific effects.

Numerous studies confirm that renewable energy reduces CO₂ emissions across different contexts. For instance, Aguir Bargaoui and Uğurlu [27,28] discovered that renewable energy adoption contributes to lowering CO₂ emissions. Uğurlu [27] identified a long-run inverse relationship between the two variables in Visegrad countries. Mukhtarov et al. [29] reported a 0.26% emissions reduction per 1% increase in renewable energy use in Azerbaijan from 1993 to 2019. Similarly, Mentel et al. [30] determined that renewable energy substantially reduces carbon emissions. Guo et al. [31] emphasized the importance of renewable energy adoption in reducing and stabilizing emissions, particularly in countries that are net exporters of carbon. In a recent study, Jahanger et al. [32] employed the method of moments quantile regression to analyze data spanning the period 1990–2020. Their findings proved that energy efficiency and the implementation of renewable energy played significant parts in reducing greenhouse gas emissions in the world’s top 10 manufacturing countries. Yang and Lo [33] highlighted the value of renewable energy and energy efficiency initiatives in China’s shift towards a carbon-neutral economy. The findings of Apergis et al. [34] are corroborated by Jahanger et al. [31], as their study proved that renewable energy negatively affects carbon emissions in Uzbekistan. Gierałtowska et al. [35] employed a fixed effects regression and a two-step GMM to examine data on 163 nations from 2000 to 2016. Their findings also demonstrate that using renewable energy reduces CO₂ emissions significantly. Similarly, Majewski et al. [36] used a two-step GMM model to investigate the relationship between renewable energy and carbon emissions in 94 middle-income countries from 2000 to 2015. They found a negative relationship between renewable energy and carbon dioxide emissions. In the same vein, Adams and Nsiah investigated the long-run relationship between renewable energy consumption and carbon emissions across 28 sub-Saharan African countries. Utilizing fully modified ordinary least squares (FMOLS) and the Generalized Method of Moments (GMM) as estimation techniques, and drawing on data from 1980 to 2014, they found evidence of a long-term association between the two variables [37]. Similarly, Raihan and Tuspekova [38] used the dynamic ordinary least squares method to estimate data on Turkey from 1990 to 2020. Their results showed a negative relation between carbon emissions and renewable energy use. In addition, Yang and Lo [33] and Ding et al. [39] emphasized the capacity of renewable energy to decrease carbon emissions. Zhang et al. [40] point out the importance of coordinated development of the energy supply network and the efficient use of renewable energy. Ding et al. [39] highlighted in their study the significance of renewable energy in multi-energy systems and the potential for optimizing energy hub planning. These studies highlight the potential of renewable energy to reduce carbon emissions; however, realizing its full impact requires well-coordinated policies and strategic planning.

While the mitigating effect of renewable energy on emissions is widely acknowledged, several scholars have drawn attention to the practical challenges and limitations involved. For example, Ang et al. [41] highlighted that the sporadic characteristics of renewable sources such as solar and wind might pose operational challenges for electricity systems, potentially limiting their effectiveness in reducing emissions. Similarly, Authors [42,43] contended that to affect global carbon emissions substantially, the shift towards renewable energy must be supported by comprehensive policy measures and modifications in behavior.

In contrast, another strand of literature reported no significant relationship between renewable energy and carbon emissions. Öztürk et al. [25], for instance, found no statistically significant relationship between renewable energy and carbon emissions in G7 countries, using the GMM-PVAR approach on data spanning from 1997 to 2019. Likewise, Saidi and Ben Mbarek [44] examined in their study the links between renewable and carbon emissions in nine developing countries, and also found no evidence of a relationship between renewable energy and carbon emissions.

2.2. Financial Development and Carbon Emissions

The current literature offers three major perspectives on the relationship between carbon emissions and financial development: positive, negative, and neutral impacts. Despite growing attention, researchers have not reached a consensus about the impact of financial development on carbon dioxide. While several academics argue that financial development is positively associated with carbon emissions, others suggest a negative relationship. A third group of scholars observed no significant relationship between financial development and carbon emissions. According to increasing empirical evidence, financial development plays a major role in driving carbon emissions [45]. The primary factor contributing to the increase in carbon emissions due to financial development is the ability of enterprises to secure funding for investments in additional manufacturing segments, expansion of production capacity, and acquisition of expensive machinery and equipment. These activities release higher levels of pollutants [46]. Charfeddine and Kahia [47] studied 24 nations in the Middle East and North Africa (MENA) region, and they proposed that financial institutions offer a borrowing system with cheap interest rates and fewer obstacles for households and investors. This, in turn, leads to higher levels of energy consumption and increased intensity of carbon emissions. The findings of Tao et al. [48] are consistent with the study conducted by Charfeddine and Kahia [47], which demonstrated that financial development had a substantial role in increasing carbon emission intensity. The findings of both Authors [47,48] confirm Akan’s [49] proposition of investigating the relationship between financial development and carbon emissions by considering the role of energy use as a mediator. Their analysis, based on panel data from G20 nations between 1999 and 2019, revealed a significant positive correlation between financial development and carbon dioxide emissions. Moreover, the relationship was found to follow an inverted U-shaped pattern when incorporating the quadratic term of financial development. However, the study also found that the combination of financial development and effective governance can significantly reduce carbon emissions [46]. Similarly, Habiba et al. [50] observed a positive correlation between financial development and carbon emissions in low-income countries, further emphasizing the context-dependent nature of this relationship.

Financial development can reduce carbon emissions by enabling businesses to allocate funds towards new technologies, introducing sophisticated technology, transferring expertise, and other methods that enhance energy performance and minimize carbon emissions. Additionally, it can incentivize governments and businesses to invest in environmentally conscious initiatives and carbon-neutral machinery by providing budget-friendly capital, thereby improving the environment’s health [46]. Ren et al.’s [51] long-run analysis showed that financial development has a substantial impact on reducing carbon emissions. However, there was no significant association detected in the short run. Authors [52,53] discovered a significant correlation between financial development and decreased carbon emissions in China. Abbass et al. [18] examined the role of financial development in promoting environmental sustainability across the Next-11 economies during the period 1990–2022, employing Common Correlated Effects Mean Group (CCEMG) and Augmented Mean Group (AMG) estimation techniques. Their findings indicated that both financial development and CO₂ emissions increased by 1.25%, suggesting a complex relationship. Financial sectors facilitate industries’ convenient access to capital from the banking sector, leading to reduced emissions. In a study of 35 sub-Saharan African countries, Aluko and Obalade [54] used the augmented mean group estimator on data from 1985 to 2014 and found a negative impact of financial development on carbon emissions.

A small body of empirical studies has found no statistically significant relationship between financial development and carbon emissions. For instance, Omri et al. [55] used the system-GMM methodology to analyze the influence of financial development on carbon emissions in 12 countries in the Middle East and North Africa (MENA) region from 1990 to 2011. Their findings suggested that there is no significant impact of financial development on carbon emissions. Similarly. Acheampong et al. [56] found that in countries with independent financial systems, the overall growth of the financial markets and their sub-indicators did not significantly impact the level of carbon emissions. Likewise, Jamel and Maktouf [57] could not identify any causal connection between the expansion of the European financial sector and carbon emissions. The results of this analysis suggest that the relationship between financial development and carbon emissions is complex and influenced by a multitude of contextual factors. The nature and intensity of the correlation depend on the indicators and conditions examined. These findings point out the need for additional research to improve our understanding of the dynamic interactions and causal mechanisms linking financial development and carbon emissions.

2.3. Industrialization and Carbon Emissions

The impact of industrialization on CO₂ emissions is predominantly positive, though some studies report insignificant effects. Li et al. [58] examined a comparable issue in China, arguing that industrialization has a significant adverse impact on air quality. Their study analyzed data from 2003 to 2017 across 31 areas in China to assess the impact of industrialization on the country’s environment. The results indicated that even moderate levels of industrial activity can have a detrimental effect on environmental quality. Similarly, Lu and Bae [59] also found that industrialization significantly influenced China’s CO₂ emissions. For Pakistan, Ullah et al. [60] used a non-linear ARDL on data from 1980 to 2018 and found that industrialization increased carbon emissions in both the long and short run. Mental et al. [30] evaluated the impacts of industrialization and the renewable energy sector on greenhouse gas emissions, focusing specifically on CO₂ emissions in Europe and Central Asia. They employed a two-step GMM method on a sample of 48 nations from 2000 to 2018. Their findings demonstrated that the process of industrialization had a positive impact on Carbon emissions. Appiah et al. [61] used AMG, CCEMG, and DCCEMG methods to empirically investigate the relationship between industrialization and carbon emissions in sub-Saharan African countries. They found a positive, insignificant relationship between industrialization and carbon emissions. Using yearly data from 1968 to 2014, Mahmood et al. [62] examined how Saudi Arabia’s industrialization affected the country’s CO₂ emissions. The results indicated that industrialization had a detrimental impact on the environment, particularly through its inelastic influence on emissions. Notably, the environmental impact was more pronounced during periods of industrial expansion than during contractions, suggesting an asymmetrical effect. Raihan and Tuspekova [38] also found that from 1990 to 2020, industrialization increased carbon emissions in Turkey. The effects of industrialization on different countries’ economies are conditional on the preexisting structure of those economies. When discussing the impact of industrialization on emissions, the consequences are influenced not only by the overall economic situation of a country but also by the specific economic characteristics of that country. In their study, Liu et al. [63] proved a positive connection between industrialization and environmental degradation, indicating that increases in industrial activity tend to exacerbate environmental deterioration. Moreover, the positive relationship between energy use and the economic cycle may increase the environmental effect of industrialization on middle-income countries. Li et al. [64] prove that energy intensity in emerging economies is affected by business cycles, as increased industrial activity during growth periods leads to increased emissions. The association highlights the role of economic changes along with industrialization in emissions, especially in middle-income countries passing through rapid development phases.

2.4. Mechanisms Linking Renewable Energy, Financial Development, Industrialization, and CO₂ Emissions

Renewable energy use, industrialization, and financial development affect CO₂ emissions differently and may exhibit variations across middle-income economies. Boya-Lara [65] and Ang et al. [41] emphasized that renewable energy can curtail emissions, but only when the grid infrastructure and policies are complementary. The intermittency of solar and wind sources constrains reductions without energy storage [41]. Recent research indicates renewable investment can decouple growth and emissions with the help of green technologies, although such benefits might be negated by rebound effects associated with increased energy demand [66,67].

Financial development provides channels to advance lower-carbon projects through green bonds and risk management mechanisms, fostering sustainability [51]. In less developed systems, it can potentially increase the emissions in the initial phase during fossil fuel financing, but a more developed system transitions to carbon-friendly effects [68]. The integration of fintech contributes to improving green investments, making green finance interventions less emissions-intensive in emerging economies [69,70].

Industrialization accelerates emissions due to the energy-intensive production that generates a carbon lock-in or the dependence on fossils [71]. Rapid urbanization increases this effect, but green policies can minimize it. In China, even renewable energy is not able to compensate for the carbon intensity of industrial output [22]. Renewables and efficiency in high-tech industries cut down emissions [9].

There is heterogeneity: upper-middle-income states use developed financial and industrial infrastructures, and lower-middle-income ones have capital and technology limitations, undermining a renewable effect [72]. Such mechanisms shape policy toward low-carbon transitions.

2.5. Literature Gaps

Examining the existing research in the previous paragraphs reveals several evident gaps. The current studies have not investigated the combined effects of renewable energy use, financial development, and industrialization on CO₂ emissions, particularly in middle-income countries. Furthermore, the application of advanced econometric techniques, such as the GMM-PVAR model, remains limited in this research domain. These shortcomings may partly explain the inconclusive and fragmented findings prevalent in the current body of carbon emissions research. Therefore, this study seeks to address these gaps by offering a more comprehensive and methodologically robust analysis, thereby enhancing the relevance of its findings for policymaking in middle-income economies.

3. Methods and Data Analysis

3.1. Data

This study investigates the effects of renewable energy consumption, financial development, and industrialization on carbon emissions in middle-income countries, using data from the World Bank’s World Development Indicators. The analysis covers the period from 2002 to 2020, selected based on data availability and to capture recent trends in economic and environmental dynamics within these economies. Our study covers 71 middle-income countries, consisting of 39 lower-middle-income countries and 32 upper-middle-income countries, as classified by the World Bank. India, Vietnam, and Kenya represent lower-middle-income countries, whereas China, Brazil, and South Africa are examples of upper-middle-income countries (Table S1). Carbon emissions, the dependent variable, are measured as metric tons per capita to capture the environmental impacts of economic activities. The independent variables include renewable energy consumption (measured as a percentage of total final energy consumption), financial development (measured as domestic credit to the private sector as a percentage of GDP, and industrialization (measured by industry value added as a percentage of GDP. This study includes two control variables: GDP per capita expressed in constant 2015 trillion US dollars, and political stability, which is an evaluation of the quality of governance and the lack of violence or acts of terrorism. Table S1 provides detailed definitions and data sources for all variables. The dataset comprises 19 years of annual data, collected on the 6 variables, across 71 middle-income countries, resulting in a total of 1349 observations. This extensive panel dataset offers a robust foundation for capturing the dynamic interactions between the variables in the context of middle-income economies. All data analysis will be done using Stata 17.

3.2. Model Specification

The main objective of the model is to analyze panel data. Building upon the theoretical foundations and empirical insights discussed earlier, particularly the contributions of [73,74], the model is specified as follows.

$${CO}_{2it}=f\left({RE}_{it},{FD}_{it},{IND}_{it},{GDP}_{it,}{PS}_{it}\right)$$

(1)

The panel data model takes the following form.

$${CO}_{2it}={\mathrm{\beta }}_0{RE}_{it}+{\mathrm{\beta }}_1{FD}_{it}+{\mathrm{\beta }}_2{IND}_{it}+{\mathrm{\beta }}_3{GDP}_{it}+{\mathrm{\beta }}_4{PS}_{it}+{\mathrm{\mu }}_{it}$$

(2)

where i is the cross-sections and t time of the study (2002 to 2020). ${\mathrm{\beta }}_0$ is the intercept while ${\mathrm{\beta }}_1$, ${\mathrm{\beta }}_2$, ${\mathrm{\beta }}_3$, and ${\mathrm{\beta }}_4$ are the coefficients of RE, FD, IND, GDP, and PS, respectively. μ represents the error term.

3.3. Estimation Procedure

The empirical analysis is presented in a systematic sequence to ensure robustness. The procedure begins with tests for cross-sectional dependence, stationarity, and cointegration. Subsequently, the GMM-PVAR model is estimated under the assumption that all variables are potentially endogenous, consistent with prior GMM-PVAR applications [73,74,75]. The panel Granger causality test follows the GMM-PVAR estimations to act as a secondary check of the directionality of relationships among variables [73,74,75].

3.3.1. Cross-Sectional Dependence

Testing for cross-sectional dependence (CD) is a critical initial step in panel data analysis, as it guides the choice of suitable econometric methods. The issue of CD arises when countries in a panel exhibit a high degree of interdependence, often stemming from international financial and economic integration [76]. This dependence Such dependence may result from unobserved error components, and the failure to recognize it may produce biased, inefficient, or inconsistent estimates [77]. To overcome this, we used the commonly used Pesaran CD test, which is appropriate when T (Time) is smaller than N (cross-sections) [76]. The null hypothesis of the test posits the absence of cross-sectional dependence in the panel. The CD test is expressed in Equations (3) and (4):

$$CSD=\sqrt{{\displaystyle \frac{2T}{N(N-1)}}}\left({\sum }_{i=1}^{N-1}{\sum }_{g=t+1}^N{\widehat{\theta }}_{ig}\right)~N\left(\mathrm{0,1}\right)i,g$$

(3)

$$B=\sqrt{{\displaystyle \frac{2T}{N(N-1)}}}\left({\sum }_{i=1}^{N-1}{\sum }_{g=t+1}^N{\widehat{\theta }}_{ig}\right)\left[{\displaystyle \frac{\left(T-K\right){\widehat{\theta }}_{ig}^2-\left(T-K\right){\widehat{\theta }}_{ig}^2}{Var\left(T-K\right){\widehat{\theta }}_{ig}^2}}\right]$$

(4)

3.3.2. Panel Unit Root Tests

The stationarity of variables in panel data was investigated using two generations of unit root tests. First-generation tests, which consider cross-sectional independence, are applicable only in the absence of cross-sectional dependence. Second-generation tests, on the other hand, are required when such dependence is present [78]. In this analysis, we employed the second-generation Cross-Section Augmented Dickey–Fuller (CADF) and Cross-Section Im-Pesaran-Shin (CIPS) tests, as proposed by Pesaran [79], to deal with the issue of cross-section dependence efficiently. These tests have been chosen due to their asymptotic properties that make them more robust and dependable in establishing the order of integration of the data [78]. The Pesaran CADF panel stationary test is expressed below in Equation (5).

$$∆Y_{it}={\alpha }_i{+\beta }_i{y}_{i,t-1}+{\delta }_i{\widehat{y}}_{t-1}+{\sum }_{d=0}^p{\rho }_{id}∆{\overline{y}}_{t-d}{\sum }_{d=0}^p{\varphi }_{id}∆y_{i,t-d}+{\epsilon }_{it}$$

(5)

where $∆{\overline{y}}_{t-d}$ and $∆y_{i,t-d}$ represent the averages and the first difference in the cross-section.

Equation (6) presents the CIPS test.

$$CIPS=N^{-1}{\sum }_{i=1}^N{CADF}_i$$

(6)

3.3.3. Westerlund Cointegration Test

Once the order of integration of the variables is established, it is essential to assess the existence of a long-run relationship among them. For this purpose, we applied the Westerlund test, a cointegration test, a second-generation approach that considers both the cross-sectional dependence and slope heterogeneity [80]. Unlike conventional tests, the Westerlund test does not rely on restrictive common factor assumptions [80]. The Westerlund cointegration test equation is expressed as follows:

$$\mathrm{\Delta }{X}_{i,t}={\beta }_i{d}_i+{\delta }_i\left(X_{i,t-1}-{\eta }_i^{’}{Y}_{i,t-1}\right)+{\sum }_{j=1}^q{\delta }_{i,j}\mathrm{\Delta }{X}_{i,t-j}+{\sum }_{j=0}^q{\lambda }_{i,j}\mathrm{\Delta }{Y}_{i,t-j}+{\epsilon }_{i,t}$$

(7)

The Westerlund test generates four statistics: two group means (Gt and Ga) and two panel (Pt and Pa) statistics. The group-mean tests (Gt and Ga) evaluate the alternative hypothesis that at least one cross-section exhibits cointegration, whereas the panel tests (Pt and Pa) test whether the entire panel is cointegrated. These statistics are derived from the error-correction terms and provide strong evidence of long-run relationships among the variables. Gt, Ga, Pt, and Pa statistics are formally expressed in the following equations.

$${\mathrm{G}}_{\tau }={\displaystyle \frac{1}{\mathrm{N}}}\sum _{i=1}^N{\displaystyle \frac{{\eta }_{\mathrm{i}}}{\mathrm{S}.\mathrm{E}.\left({\hat{\eta }}_{\mathrm{i}}\right)}}$$

(8)

$${\mathrm{G}}_{\mathrm{a}}={\displaystyle \frac{1}{\mathrm{N}}}\sum _{i=1}^N{\displaystyle \frac{\mathrm{T}{\eta }_{\mathrm{i}}}{\eta {ˊ}_{\mathrm{i}}\left(1\right)}}$$

(9)

$$\mathrm{\Delta }{\mathrm{P}}_{\tau }={\displaystyle \frac{{\hat{\eta }}_{\mathrm{i}}}{\mathrm{S}.\mathrm{E}.\left({\hat{\eta }}_{\mathrm{i}}\right)}}$$

(10)

$${\mathrm{P}}_{\mathrm{a}}=\mathrm{T}\hat{\eta }$$

(11)

3.3.4. GMM-PVAR Model

We have implemented the GMM-PVAR model to explore the dynamics between the variables in our study. The GMM-PVAR model is a highly sophisticated econometric tool used to examine dynamic interactions between two or more endogenous variables in panel data contexts. It is based on the standard Panel Vector Autoregression (PVAR) and combines the General Method of Moments (GMM) to overcome common time series problems with endogeneity, autocorrelation, and fixed effects [81]. The GMM-PVAR model is an improvement proposed by Sigmund and Ferstl [81], which introduces lagged endogenous variables, strictly exogenous variables, and weakly exogenous covariates to provide an effective estimation of structural relationships [82].

It addresses the shortcomings of traditional panel data methods, such as Ordinary Least Squares (OLS), which include the presence of a Nickell bias, which is still present when the number of cross-sectional units becomes large $\left(N\to \infty \right)$ and the time (T) is fixed [83].

The GMM-PVAR model’s structural form is specified as:

$$m_{i,t}=p_i+{\sum }_{j=1}^k{G}_j{m}_{i,t-j}+W{\phi }_{i,t}+{U\lambda }_{i,t}+{\epsilon }_{i,t}$$

(12)

In Equation (12) $m_{i,t}$ is the endogenous factor with time t, $m_{i,t-j}$ determines the lagged value of the endogenous variables, $p_i$ is an identity of size (n × n), and G, W, and U are the homogeneity parameters. ${\phi }_{i,t}$ represent weakly exogenous variables, and ${\lambda }_{i,t}$, shows a vector of strictly exogenous variables that are not related to the error term, where λ = 1,..., T. Lastly, ${\epsilon }_{i,t}$, denotes the individual error term, which is assumed to be independently and identically distributed. There are two approaches to address fixed effects. The first-difference and the forward orthogonal transformation methods help mitigate the issues associated with fixed effects. However, the GMM methodology allows us to circumvent these procedures. Specifically, by retaining the transformation matrix that links differenced variables to lagged covariates, these parameters can be used as lagged regressors in the GMM estimation. Authors [73,74,75,82] specified [81] the first difference approach in the GMM modeling as follows:

$$\mathrm{\Delta }{m}_{i,t}={\sum }_{j=1}^k{G}_j\mathrm{\Delta }{m}_{i,t-j}+W\mathrm{\Delta }{\phi }_{i,t}+U\mathrm{\Delta }{\lambda }_{i,t}+\mathrm{\Delta }{\epsilon }_{i,t}$$

(13)

In Equation (13), Δ represents the first difference. Our lagged endogenous variables are CO₂, RE, FD, IND, GDP, and PS. We verified the stability of the PVAR model by examining the modulus of the eigenvalues of the computed models. The stability model can only meet the PVAR criteria with eigenvalues inside the circle. We used the method described by Andrews and Lu [84] for lag selection. The study presented three distinct measures utilizing the model and moment selection criteria (MMSC), specifically the Hannan-Quinn information criterion (MMSC-HQIC) and the Bayesian information criterion (MMSC-BIC). In our investigation, lag length determination was examined based on MMSC-BIC [73,74,75], where a lag length of q = 1 was used for our models.

Endogeneity arises when the error terms in econometric models are correlated with the explanatory variables, leading to biased and inefficient estimates. For example, Chinese industrial growth may contribute to emissions while simultaneously influencing renewable energy policies, creating feedback loops in which one variable affects another over the long run. While the standard GMM approach addresses some endogeneity concerns, it does not fully capture cross-country interactions, such as trade or policy spillovers. Given the significance of endogeneity in panel data models, the GMM-PVAR technique provides a robust solution by constructing a system of equations in which all variables are treated as endogenous [82,85].

3.3.5. Panel Granger Causality

Our goal was to find out which factors impact Carbon emissions by conducting the panel VAR Granger causality test. We are utilizing this approach since all our chosen variables exhibit stationarity following the initial differencing. We have developed the subsequent framework for our panel VAR Granger causality test, as outlined by [86]:

$$Q_{i,t}={\delta }_0+{\sum }_{q=1}^b{\delta }_{1,j}{L}_{i,t-j}+{\sum }_{q=1}^b{\delta }_{2,j}{L}_{i,t-j}+{\theta }_i+{\epsilon }_{i,t}$$

(14)

$$D_{i,t}={\kappa }_0+{\sum }_{q=1}^b{\upsilon }_{1,j}{S}_{i,t-j}+{\sum }_{q=1}^b{\upsilon }_{1,j}{S}_{i,t-j}+{\tau }_i+{\mu }_{i,t}$$

(15)

In Equations (14) and (15), b represents the lag length, i and t represent countries and time periods. While ${\theta }_i$ and ${\tau }_i$ indicate the respective fixed effects ${\epsilon }_{i,t}$ and ${\mu }_{i,t}$. These are the white noise error components.

Furthermore, if the lagged variable Q contains information relevant to the variable D, then changes in Q may have predictive power for changes in D, and vice versa. The VAR Granger causality test utilizes chi-squared statistics to determine whether to reject the null hypothesis. The null hypothesis posits the absence of causality among the panel variables, whereas the alternative hypothesis indicates the presence of causality.

4. Results and Discussion

The preliminary analysis is a crucial and fundamental stage before an empirical study. It helps make well-informed decisions about the course of the research and provides the foundation for additional analysis. This will give further insights into the attributes and quality of the data. Our initial statistical analysis begins with the summary statistics of our data. Table S2 presents the descriptive statistics of the variables examined in both panels, including the Mean, Median, Maximum, Minimum, standard deviation (Std. Dev.), Skewness, and Kurtosis. As shown in Table S2, the mean values of CO₂ emissions are notably higher in the upper middle-income countries (3.615) compared to lower–middle–income countries (1.286). Average renewable energy use is higher in higher and lower-middle-income countries. As anticipated, upper-middle-income nations exhibit higher financial development, GDP, and industrialization levels. The data also reveal significant heterogeneity between both panels’ minimum and maximum values. Most of the variables exhibit positive skewness, indicating the presence of extremely high values, except for renewable energy consumption in lower-middle-income countries, which shows a slight negative skewness (−0.109). Excess kurtosis (over 3) is observed in all variables, especially industrialization in upper-middle-income countries (6.620) and GDP in lower-middle-income countries (7.255), showing a distribution that is highly skewed and, thus, not normal. To address this and stabilize variance, all variables were transformed into their natural logarithm form for further analysis.

Table 1. Pearson Correlation and VIF results.

	CO2	RE	FD	GDP	IND	PS	VIF
Upper Middle-Income Countries
CO2	1
RE	−0.697	1					1.49
FD	0.384	−0.265	1				1.31
GDP	0.719	−0.550	0.448	1			2.19
IND	0.579	−0.339	0.325	0.607	1		1.79
PS	0.029	0.044	0.261	0.174	0.369	1	1.26
Lower Middle-Income Countries
CO2	1
RE	−0.552	1					1.93
FD	0.141	0.055	1				1.56
GDP	0.417	−0.167	0.064	1			3.04
IND	0.467	−0.118	0.231	0.436	1		1.69
PS	−0.096	0.241	0.337	0.146	0.255	1	1.27

presents Pearson correlation and variance inflation factors (VIF) results. Renewable energy (RE) was negatively correlated to CO₂ in both upper and lower-middle-income countries. RE had a consistent negative relationship with CO₂ emissions in both income groups; however, the relationship was more substantial in upper-middle-income countries (−0.697) compared to lower-middle-income countries (−0.552). This implies that the abatement potential of CO₂ is relatively higher in more economically developed middle-income economies, potentially due to superior technological capacities and institutional systems. Financial development (FD), GDP per capita, and industrialization (IND) were positively correlated with CO₂ emissions in both groups. GDP showed stronger correlations with CO₂ emissions in the upper middle-income countries (0.719) than in lower middle-income countries (0.417), revealing that the income-emissions association becomes stronger with economic progress. In contrast, Political stability (PS) exhibited opposing trends, showing a weakly positive correlation (0.029) in upper-middle-income nations, while displaying a slight negative correlation (−0.096) in lower-middle-income nations. Jianguo et al. [87] noted that multicollinearity adversely affects regression results when the Variance Inflation Factor (VIF) for variables exceeds a threshold value of 10. As shown in Table 1, all VIF values are below the threshold, with the maximum of 3.04. The majority of VIF values were under 2.0, indicating the absence of severe multicollinearity and thus supporting the reliability of the subsequent GMM-PVAR estimations.

The results of the cross-sectional dependence test are displayed in

Table 2. Cross-sectional Dependence Test Results.

	Upper Middle-Income Countries	Lower Middle-Income Countries
CO2	133.98 ***	47.52 ***
RE	34.71 ***	4.12 ***
FD	106.38 ***	49.77 ***
GDP	140.62 ***	57.12 ***
IND	119.83 ***	47.12 ***
PS	78.48 ***	37.32 ***

*** signifies 1% level of significance.

below. As stated in the methodology section, we used the Pesaran cross-sectional dependence test [88]. The CD test serves as the fundamental basis for all other econometric approaches utilized in this research. Based on Table 2, we reject the null hypothesis of no cross-sectional interdependence. This confirms the presence of cross-sectional dependence among all the variables analyzed in this study in Upper Middle-Income and Lower Middle-Income Countries. Given this finding, estimation techniques capable of addressing cross-sectional dependence are required. Accordingly, examining the second-generation tests in the subsequent parts of the econometric analysis is necessary.

We performed the second-generation CIPS and CADF unit root test, as specified by [79], after determining the presence of cross-sectional dependence among our variables. The results are displayed in

Table 3. CIPS and CADF Unit Root Tests results.

	CIPS		CADF
	Level	First Difference	Level	First Difference
Upper Middle-Income Countries
CO2	−1.596	−3.413 ***	−1.527	−2.672 ***
RE	−2.076	−3.195 ***	−1.425	−3.419 ***
FD	−2.133	−3.329 ***	−1.882	−2.396 ***
GDP	−1.824	−2.986 ***	−1.532	−2.193 ***
IND	−1.879	−4.404 ***	−2.003	−2.891 ***
PS	−2.021	−4.012 ***	−1.780	−3.135 ***
Lower Middle-Income Countries
CO2	−1.772	−4.024 **	−1.356	−2.945 ***
RE	−1.625	−3.680 ***	−1.176	−3.079 ***
FD	−1.964	−3.509 ***	−1.800	−2.230 **
GDP	−1.978	−3.126 ***	−1.337	−2.384 ***
IND	−1.310	−4.674 ***	−1.350	−3.342 ***
PS	−1.821	−3.902 **	−1.602	−2.794 ***

*** and ** signifies 1% and 5% statistical levels of significance, respectively.

below. CIPS and CADF test outcomes demonstrated that all the variables in both upper and lower-middle-income countries were non-stationary at their level. However, after first differencing, all variables became stationary, confirming integration of order one. This transformation eliminated unit roots across the dataset. Moreover, all first-differenced variables exhibited a high level of statistical significance, validating the reliability of the unit root test results. Therefore, we can now proceed to test for cointegration.

The Westerlund cointegration test results (Table S3) confirmed a long-run relationship among all variables. The test statistics Gt, Ga, Pt, and Pa were all statistically significant, indicating cointegration across both panels. This suggests a stable long-term equilibrium among CO₂, RE, FD, IND, GDP, and PS in both upper- and lower-middle-income countries.

The GMM-PVAR estimation results presented in

Table 4. GMM-PVAR estimation results.

	CO2	RE	FD	GDP	IND	PS
Upper Middle-Income Countries
L1.CO2	−0.054	−0.135	0.869 ***	−0.647 ***	0.602 ***	−0.568 **
L1.RE	−0.013	0.299 ***	0.136	−0.126	0.039	−0.023
L1.FD	−0.087***	−0.022	0.947 ***	−0.035	−0.064 ***	−0.114 ***
L1.GDP	0.126 ***	0.074	−0.056	0.948 ***	0.064	0.106
L1.IND	2.030 **	0.104	0.210	−0.004	0.493 ***	−0.330 **
L1.PS	0.083 **	−0.079 *	−0.036	0.113 ***	0.074	0.877 ***
Lower Middle-Income Countries
L1.CO2	0.105	−0.186	0.129 *	−0.558 *	0.041	0.076
L1.RE	0.047	0.895 ***	0.008	−0.090	0.054	−0.058
L1.FD	−0.069 **	−0.046	0.904 ***	−0.062	−0.065 ***	−0.116 **
L1.GDP	0.081	0.164 *	−0.218	0.967 ***	−0.021	0.009
L1.IND	0.196	0.013	0.182	0.002	0.242	−0.383 *
L1.PS	0.184 **	−0.079	−0.211	0.089	0.076	0.828 ***

***, **, and * signifies 1%, 5%, and 10% statistical levels of significance, respectively.

reveal the presence of long-run equilibrium relationships among renewable energy consumption, financial development, industrialization, and carbon emissions of both upper and lower middle-income countries.

4.1. Upper Middle-Income Countries

In upper-middle-income economies, long-run relationships indicate a challenge to conventional assumptions of environmental economics. Contrary to expectations, renewable energy use (L1.RE) showed a negative but non-significant effect on CO₂ emissions (−0.013, p > 0.10). This suggests renewable energy’s carbon abatement potential may be limited in developed upper-middle-income economies may be limited. This observation is consistent with Wang et al. [22], who assert that renewable energy’s capacity to reduce emissions weakens at higher development levels due to increased energy demand and industrial complexity.

The financial development–emissions nexus conveyed an inverse relationship, as financial development (L1.FD) was found to have a negative impact on CO₂ emissions (−0.087, p < 0.01). This finding confirms the growing evidence that green financing and environmental technology adoption rely on established financial systems in upper-middle-income countries. Similarly, Sethi et al. [89] observed that financial development significantly reduced CO₂ emissions across 25 developing countries. They attributed this outcome to the more efficient allocation of capital toward cleaner and environmentally friendly technologies facilitated by advanced financial systems.

The most prominent result is the strong positive relationship between industrialization (L1.IND) and CO₂ emissions (2.030, p < 0.05), indicating that industrial expansion substantially drives carbon output in the long run. Gopalakrishnan and Miller [71] pointed out that upper-middle-income nations experience “industrial carbon lock-in”. In such cases, the structural legacy of industrial society produces path dependence effects that slow down decarbonization progress despite the acquisition of renewable energy sources. Political stability (L1.PS) had a positive correlation with CO₂ emissions (0.083, p < 0.05), which implies that a stable political climate in upper-middle-income countries facilitates sustained industrialization, which in turn temporarily elevates emissions. This paradox aligns with the findings of Hacıimamoğlu and Sungur [90], who observed a strong connection between pro-growth policies and political stability, as the latter often underpins pro-growth policies that prioritize economic expansion over environmental constraints.

4.2. Lower Middle-Income Countries

The panel results for lower-middle-income countries revealed distinct long-run dynamics, supporting the hypothesis of systematic variation in environmental-economic relationships across development pathways. Renewable energy exhibited a non-significant positive impact on CO₂ emissions (0.047, p > 0.10); however, its strong autoregressive persistence (0.895, p < 0.01) suggests that renewable energy investments in these economies exert growing influence over time. In contrast,

Financial development (L1. FD) showed a negative and significant effect on CO₂ emissions (−0.069, p < 0.05), although the magnitude of this effect was smaller than that observed in upper-middle-income countries. Yadav et al. [91] found that energy efficiency gains associated with financial development are a more significant driver of emission patterns than the widespread deployment of renewable energy sources. GDP growth (L1.GDP) in lower-middle-income countries exhibited a moderate positive relationship with renewable energy use (0.164, p < 0.10), suggesting that economic growth may exert an indirect influence that, over time, facilitates the transition toward a cleaner energy system. These findings align with the study of Kamah et al. [92], whose “growth-first, clean-later” paradigm characterizes the fundamental development logic prevailing in emerging economies.

Political stability (L1.PS) positively impacts CO₂ emissions (0.184, p < 0.05), indicating that stable political regimes in lower-middle-income countries may enable economic expansion that elevates short-term emissions, while concurrently increasing the capacity to enhance the environment in the long term. The contrasting findings between the income-based subgroups demonstrate that the environmental-economic nexus follows distinct pathways at different development stages.

The insignificant effect of renewable energy use on CO₂ emissions in both middle-income country groups points to decarbonization obstacles. Renewable energy growth (180.79 GW non-fossil capacity in India by 2023 [13]) is an effective strategy, but there are rebound effects, grid stability concerns, and coal-intensive industrial systems. Rebound effects contribute to more energy use due to renewable energy cost reductions, especially in lower-middle-income countries with lopsided access and upper-middle-income countries with heavy industrial demand [93,94,95]. Dawn et al. [96] attribute grid limitations, such as the predominance of coal in China as a reliable baseload, to factors that slow renewable energy integration, including aging infrastructure and regulatory obstacles. Heavy industries in upper-middle-income countries perpetuate ‘carbon lock-in’, fighting renewable energy substitution [71]. Lower-middle-income countries are threatened with the same patterns without structural changes

The observed heterogeneous impacts across income groups can be attributed to structural and developmental differences. In upper-middle-income economies, more developed industries and an established industrial base increase emissions because they are known to use carbon-intensive industries (coefficient 2.030, p < 0.05). The strong positive effect of industrialization on CO₂ emissions in upper-middle-income economies is explained mainly by heavy manufacturing industries, like as steel, cement, and chemicals, in countries like China [14]. These industries depend on coal because it provides high-density, dependable energy, which forms path dependencies to maintain high emissions despite increases in renewable energy. Brazil demonstrates a preventive track, through its biofuel-industrial synergy, decreasing the need to use fossil fuels [15]. In lower-middle-income countries, like India and Vietnam, the impact of industrialization is insignificant, possibly because these countries do not have as deeply established systems of industry and thus have the possibility of cleaner industrialized futures. The higher emission-reducing impact of financial development in upper-middle-income countries (−0.087) than in lower ones (−0.069) can be explained by more advanced financial systems that facilitate the provision of green investments in the former. The differences highlight the importance of context-specific policies, especially suited to the economic and environmental context of each subgroup [72].

The Granger causality tests (

Table 5. Granger Causality test results.

		CO2	RE	FD	GDP	IND	PS
Upper Middle-Income Countries	CO2		0.419	9.533 ***	18.630 ***	10.359 ***	5.659 **
	RE	0.070		2.014	2.732	0.667	0.062
	FD	15.741 ***	1.152	0.53	1.433	11.050 ***	11.889 ***
	GDP	8.690 ***	2.216			2.183	2.828
	IND	4.610 **	0.766	1.574	0.003		6.162 ***
	PS	4.15 **	2.836	0.234	7.413	58.131
Lower Middle-Income Countries	CO2		0.217	3.354 *	3.377 *	0.014	0.032
	RE	0.466		0.003	0.538	1.130	0.211
	FD	4.892 **	1.837		1.302	10.740 ***	4.545 **
	GDP	1.174	3.704 *	2.293			0.008
	IND	2.368	0.005	0.775	0.000	0.126	3.009 *
	PS	4.297 **	0.513	1.276	0.097	0.408

***, **, and * signifies 1%, 5%, and 10% statistical levels of significance, respectively.

) reveal no causal relationship between renewable energy use and CO₂ emissions in either income group. However, a bidirectional causal relationship exists between financial development and CO₂ emissions across both income groups. A reciprocal relationship linked carbon emissions with GDP in upper-middle-income nations. Whereas in lower-middle-income countries, the causality was unidirectional, running from CO₂ emissions to GDP.

Furthermore, industrialization and CO₂ emissions demonstrated a bidirectional causality in upper-middle-income nations, while no significant causal link was observed in lower-middle-income counterparts. Political stability demonstrated a bidirectional causative relationship with CO₂ emissions in upper-middle-income nations, while. It only showed a unidirectional causal relationship in lower-middle-income countries.

Following GMM-PVAR estimation and Granger causality testing, we proceeded to assess the stability of the model. The eigenvalue stability condition confirms all modulus values lie within the unit circle (|λ| < 1), indicating that the GMM-PVAR model is dynamically stable. As visually confirmed in Figure 3 and Figure 4, the comprehensive outcomes of the stability evaluation affirm the robustness, reliability, and replicability of the core model estimations.

4.3. Robustness Check

To ensure robustness, we also employed conventional econometric methods, FMOLS and Dynamic ordinary least squares (DOLS), to compare with the results derived from the GMM-PVAR technique. The estimated outcomes from these traditional approaches are presented in Table S4. Overall, the conventional methods yielded contradictory results, particularly regarding the effects of RE, FD, and IND on Carbon emissions. The main purpose of presenting the findings of conventional techniques is to emphasize the significant differences between the results obtained from advanced and traditional econometric methodologies. Notably, conventional techniques fall short in addressing endogeneity and cross-sectional dependence. As a result, their findings may be unreliable, ambiguous, and potentially biased, thereby limiting their usefulness in guiding effective policy formulation.

5. Conclusions

Mounting global decarbonization imperatives necessitate a deeper understanding of how middle-income economies navigate the development–emission nexus through key drivers like renewable energy, financial systems, and industrial transformation. This paper provides new empirical insights into the interplay between environmental and economic factors in middle-income economies. In this paper, the effects of renewable energy use, financial development, and industrialization on CO₂ emissions in middle-income economies have been analyzed using the GMM-PVAR model. By analyzing data from 71 middle-income countries over the period 2002–2020 and distinguishing between upper- and lower-middle-income subgroups, the study provides valuable evidence of the heterogeneous nature of environmental–economic relationships across different stages of development

The GMM-PVAR estimation results indicated the existence of highly heterogeneous long-run relations across income groups, challenging the assumption of uniform environmental-economic dynamics within middle-income economies. In upper-middle-income countries, renewable energy consumption showed no effect on CO₂ emissions, suggesting that carbon abatement potential from clean energy technologies may be constrained by rising aggregate energy demand and industrial sophistication. These findings deviate from conventional expectations and underscore the necessity of implementing complementary policies alongside the promotion of renewable energy to achieve substantial emission reductions. Financial development demonstrates a negative effect on CO₂ emissions in both income groups, though with varying magnitudes. The coefficient of −0.087 in upper-middle-income countries indicates that mature financial systems enable greater emission reductions, likely through the more efficient allocation of capital toward green technologies and environmentally focused projects. In contrast, while still significantly negative, the magnitude of the effect was smaller (−0.069) in lower-middle-income countries. This pattern indicates that financial sector development in these economies primarily supports environmental improvement by enhancing energy efficiency rather than enabling the widespread deployment of renewable energy technologies. The industrialization–emissions nexus yielded the most pronounced findings, with the upper-middle-income countries exhibiting a high positive coefficient (2.030), indicating the strong carbon intensity of industrial growth. This finding reflects the phenomenon of industrial carbon lock-in, where legacy industrial infrastructure generates path dependencies that hinder rapid decarbonization. Conversely, industrialization showed no significant impact on CO₂ emissions in Lower-Middle-Income Countries, suggesting early-stage industrial development may enable cleaner pathways. Political stability was positively associated with emissions in both income groups, albeit likely through different mechanisms. In upper-middle-income countries, stable governance enables sustained pro-growth policies that temporarily increase emissions. Conversely, in lower-middle-income countries, political stability may create the institutional conditions necessary for economic activity that eventually enables long-term environmental improvements.

The heterogeneous effects in upper- and lower-middle-income countries also indicate the necessity of differentiated measures, with upper-middle-income economies having to cope with industrial carbon lock-in and lower-middle-income economies with financial and technological capacity building assistance.

6. Policy Recommendations

We found substantial differences in the driving factors of emissions among middle-income country subgroups, underscoring the need to develop customized policy frameworks incorporating unique structural challenges and opportunities. Although our 2002–2020 study period lags current technological advancements, the post-COVID developments add to the viability of recommendations covered in our study and emphasize the necessity to address structural obstacles identified.

6.1. Upper Middle-Income Countries

The strong positive effect of industrialization on emissions in upper-middle-income economic states indicates deeply established carbon-intensive industry structures, and such states need specifically designed interventional measures.

Upper-middle-income economies should use carbon border adjustments (CBAs) to ensure industrial competitiveness as they shift to clean production. These instruments aim to cushion domestic steel, cement, and aluminum companies that invest in low-carbon technology against unfair competition with high-carbon imports. The benefits of CBAs of trade-exposed and energy-intensive sectors would be especially significant in China and Brazil, which are major industrial economies in our sample.

The focus on funding just transitions in coal-dependent industrial regions needs to be comprehensive, with support that may include worker retraining programs, economic diversification programs, and infrastructure modernization. The fund should be a combination of both government involvement and multilateral development banks, and it should have funds specific to regions such as Shanxi Province in China or the coal mining regions of Brazil.

Since our results indicate that financial development is a more effective mitigation activity in upper-middle-income countries, these countries should adopt a similar approach and require financial institutions and major corporations to disclose how they address climate risks. This policy utilizes the current level of financial sophistication to shift capital flows to sustainable projects.

6.2. Lower Middle-Income Countries

The negative industrialization effect on lower-middle-income countries suggests that these economies still have flexibility in avoiding carbon-intensive growth pathways by being proactive in designing complementary policies.

Both grid modernization and renewable energy financing infrastructure should be prioritized in blended finance mechanisms that incorporate concessional public funds with additional resources leveraged through the private sector. The nations such as Vietnam and Indonesia need significant grid investments to match renewable capacity without compromising reliability. The mechanisms must work to achieve 70 percent of the public and 30 percent of the private funding arrangements to improve market failures in the infrastructure provision.

Green industrial parks that provide shared clean energy infrastructure and central waste management systems, and an integrated system of environmental monitoring, can lower per-firm compliance costs and allow cleaner industrial development. The current special economic zone framework in India offers a platform to extend the green industrial parks template to other lower-middle-income economies.

Technology transfer agreements with the support of climate finance schemes must center on accessing cleaner industrial technology that is not accessible via market-based mechanisms alone. These agreements must have compulsory technology localization that should ensure building domestic capacities instead of the importation of equipment to the country.

Both income groups will need international coordination mechanisms such as harmonized environmental standards, shared research initiatives, and a technology sharing nexus to avert carbon leakage and guarantee global policy effectiveness.

The differentiated approach is rooted in our empirical evidence: upper-middle-income countries need to actively decarbonize their current carbon-intensive systems, whereas lower-middle-income countries have an opportunity to build clean development foundations in the first place. Policy implementation success is linked to tailoring policy measures to the specific structural features and development limitations of each group.

7. Limitations and Future Research Directions

These limitations constrain the interpretability of our findings. First, the use of composite data for renewable energy may obscure technology-specific effects, potentially masking the distinct environmental impacts of sources such as solar, wind, hydro, and biomass. Aggregated renewable energy measures cannot tell us whether intermittent sources (solar and wind power) or baseload sources (hydro) are increasing. This aggregation is potentially why renewable energy use did not significantly affect the level of CO₂ emissions in our GMM-PVAR estimates (Table 4), since differences in implementation and efficacy across the types are not modelled. Policymakers are therefore encouraged to differentiate between these sources in future analyses to tailor more effective energy strategies. Also, the analysis is based on production-based CO₂ emissions, which may not fully capture the environmental footprint of countries with substantial trade volumes. In such cases, consumption-based emissions accounting could provide alternative insights, especially in economies that are net importers or exporters of carbon-intensive goods.

Although covering a substantial period (2002–2020), it may not fully reflect the most recent advancements in renewable energy technologies or policy shifts across countries. These recent developments could potentially alter the dynamics observed in our analysis. Therefore, future research should consider extending the dataset to incorporate the most up-to-date information, which may offer new insights and improve the robustness of policy recommendations.

The mechanisms underlying the effects of financial development on CO₂ emissions warrant further investigation to determine whether the observed effects operate through channels including green finance, improvements in capital allocation efficiency, or other structural pathways. Additionally, micro-level research, particularly studies focusing on firm-level adoption of renewable energy and corresponding emission outcomes, could yield deeper insights into the aggregate relationships established in this study.