Globalization, Financial Risk, and Environmental Degradation in China: The Role of Human Capital and Renewable Energy Use

Abstract

Amid rising climate concerns, understanding how renewable energy adoption, human capital, fossil fuel efficiency, and globalization collectively shape CO₂ emissions is crucial for unlocking pathways to a cleaner, resilient, and globally connected low-carbon future. Using China as a case study, this research investigates the drivers of CO₂ emissions, focusing on fossil fuel efficiency, renewable energy adoption, and globalization, utilizing quarterly data from 1984Q1 to 2023Q4. To ensure robust and nuanced insights, the study integrates advanced machine learning techniques alongside Quantile-on-Quantile Kernel Regularized Least Squares (QQ-KRLS) and a Modified Quantile Regression as robustness checks, capturing complex distributional dynamics often overlooked in conventional analyses. To the authors’ knowledge, this is the first empirical study examining such relationships for the case of China. The results reveal that globalization, fossil fuel efficiency, renewable energy, human capital, and financial risk all contribute to increasing CO₂ emissions. The study proposes precise policies based on the findings obtained.

1. Introduction

China is the world’s largest emitter of CO₂, accounting for nearly 35% of global carbon emissions in 2023, far outpacing other nations (https://www.iea.org/reports/co2-emissions-in-2023/the-changing-landscape-of-global-emissions, 21 July 2025). Rapid industrialization and economic expansion have fueled this surge: between 2000 and 2020, Chinese CO₂ emissions rose from ~3.4 Gt to over 10 Gt—an average annual growth rate of about 5.4%. According to China’s National Bureau of Statistics, coal made up 56% of the country’s total primary energy consumption in 2021, and coal combustion was responsible for roughly 79% of energy-sector CO₂ emissions (https://climateactiontracker.org/countries/china/2022-11-03/policies-action/, 21 July 2025). In response, the central government has implemented stronger decarbonization policies—setting goals to peak emissions before 2030 and become carbon-neutral by 2060—but current policy ratings remain “insufficient,” and emissions as of 2024 are still around 15.8 GtCO₂e, with only marginal decreases year on year [1]. At the same time, China is investing heavily in clean energy both at home and abroad. In 2024 alone, clean energy investment reached approximately CNY 6.8 trillion (USD 625–940 billion), accounting for nearly one-third of global clean energy funding and nearly two-thirds of global energy transition spending [2]. Its additions of solar PV and wind capacity in 2023 and early 2025 broke records—commissioning as much PV as the rest of the world did in all of 2022 and adding 93 GW solar and 26 GW wind in early 2025 [2]. Overseas, around 68% of its recent BRI energy investments went to solar and wind projects. Nevertheless, meeting its carbon neutrality goal remains challenging: despite early peaks in renewables, China continues to approve significant coal capacity (94 GW under construction in 2024), and policy weaknesses like regional inconsistencies in its emissions trading system hinder progress.

Despite China’s rapid growth in renewable energy capacity, with record-breaking additions in solar and wind power in 2023 and 2024, the nation is still witnessing a surge in CO₂ emissions, exacerbating environmental degradation and air quality concerns [3]. This paradox stems from China’s simultaneous expansion of coal-fired power to support industrial growth and stabilize its energy system, leading to continued high emissions despite clean energy investments [1]. However, renewables hold significant potential to help China reduce CO₂ emissions if integrated effectively into its grid and coupled with coal phase-down policies, energy storage, and grid flexibility improvements. Studies have shown that scaling up renewables while reducing coal dependence can substantially cut emissions and advance China’s carbon neutrality goals while supporting energy security [4].

Fossil fuel efficiency (FFE) refers to improving the energy output per unit of fossil fuel consumed, aiming to generate the same or higher levels of economic productivity while using less fuel and emitting fewer pollutants. It involves technological upgrades such as advanced combustion systems, waste heat recovery, and fuel-switching to less carbon-intensive fuels within the fossil fuel category (e.g., from coal to natural gas) to reduce energy intensity and emissions per output unit [5]. Increasing fossil fuel efficiency is crucial in transitional energy systems where fossil fuels still dominate, as it provides immediate opportunities for reducing CO₂ emissions without waiting for a complete renewable energy transition, thereby supporting near-term climate targets while ensuring energy security in fast-growing economies [6,7]. Empirical studies support this link; ref. [8] found that an improvement in FFE led to a reduction in CO₂ emissions in China’s industrial sector, while Lee et al. (2021) [9] reported that higher FFE correlates with lower per capita CO₂ emissions. Thus, improving FFE serves as a practical pathway to reduce CO₂ emissions and environmental degradation while supporting economic activity during the clean energy transition.

Globalization, characterized by increased cross-border trade, capital flows, and technological transfers, has a complex relationship with CO₂ emissions. On the one hand, it can lead to higher emissions due to increased industrial production, transportation, and energy consumption associated with expanded trade and economic activities [4,10]. Many developing countries, including China and India, have experienced rising emissions as they integrate into global supply chains, often relying on fossil fuels to power export-driven growth [11]. Additionally, the “pollution haven hypothesis” suggests that globalization may lead high-income countries to relocate carbon-intensive industries to countries with less stringent environmental regulations, further raising emissions in these host nations [12]. Conversely, globalization can also facilitate CO₂ emissions reduction by enabling technology transfer, diffusion of cleaner production processes, and foreign direct investment in energy-efficient and renewable energy projects [13]. It can help countries access advanced low-carbon technologies, improve energy efficiency, and adopt best practices in environmental management, thereby reducing their carbon footprint while maintaining economic growth. For example, evidence from [14] suggests that globalization contributes to environmental improvements in emerging economies when accompanied by green technology inflows and strong environmental governance. Therefore, the interaction between globalization and CO₂ emissions depends on the balance between increased economic activity and the adoption of cleaner technologies facilitated by global integration.

Human capital, characterized by the education, skills, and health status of a population, has a complex relationship with CO₂ emissions. On the one hand, higher levels of human capital can initially contribute to increased emissions by driving economic growth, industrial expansion, and energy consumption as skilled labor enhances productivity and supports higher output levels [15]. Many developing countries, including China and India, have witnessed rising emissions alongside improvements in human capital as they pursue industrialization and urbanization pathways. Additionally, the “scale effect” implies that as human capital expands economic activities, energy use may rise, thereby increasing emissions if fossil fuels dominate the energy mix [16,17]. Conversely, human capital can play a crucial role in reducing CO₂ emissions by facilitating the adoption of cleaner technologies, improving energy efficiency, and fostering environmental awareness. A well-educated workforce can enable innovation in renewable energy and low-carbon technologies, enhance the implementation of sustainable practices, and support the transition towards greener economic structures. For example, evidence from emerging economies suggests that human capital development contributes to environmental improvements when coupled with strong institutional frameworks and green innovation policies [18]. Therefore, the interaction between human capital and CO₂ emissions depends on the balance between growth-driven energy demand and the potential to drive low-carbon transitions through knowledge and technology diffusion.

Based on the information above, this study examines the drivers of CO₂ emissions in China by asking the following:

(a)

How does China’s economic growth and globalization interact with emissions trends?

(b)

What is the effect of fossil fuel efficiency (FFE) on emissions trends in China?

(c)

How does human capital impact emissions trends in China?

(d)

What is the association between renewable energy and CO₂ emissions?

(e)

How can China achieve carbon neutrality?

These questions are essential within the framework of the United Nations Sustainable Development Goals (SDGs), particularly SDG 7 (Affordable and Clean Energy), SDG 9 (Industry, Innovation, and Infrastructure), SDG 12 (Responsible Consumption and Production), and SDG 13 (Climate Action). Investigating these drivers will provide evidence on how China can balance economic growth with environmental sustainability, offering insights for policymakers aiming to advance low-carbon pathways while meeting development goals. The study also aims to contribute to global discussions on sustainable industrialization and climate mitigation by demonstrating how fossil fuel efficiency and renewable energy investments can jointly facilitate CO₂ emissions reductions in emerging economies.

This study makes significant contributions to the ongoing literature on climate and energy policy in China. First, it pioneers an investigation into the impact of fossil fuel efficiency (FFE) on CO₂ emissions in China, offering fresh empirical insights that can reshape policy discussions on transitional decarbonization pathways. Second, unlike conventional approaches, this study considers the entire distribution of CO₂ emissions and its drivers by employing the recently introduced Cross-Quantile Regression method by [19], allowing for a nuanced understanding of how these relationships behave across low, median, and high emission levels. Third, by incorporating critical but often overlooked factors such as globalization and financial risk, the study enriches the discourse on CO₂ emissions by highlighting the multidimensional influences shaping China’s carbon trajectory. Collectively, these contributions position the study as a valuable resource for policymakers and researchers seeking targeted, quantile-based strategies to advance China’s emission reduction goals while maintaining sustainable economic growth.

The next section discusses the theoretical framework and the literature; Section 3 presents the data and methodology; Section 4 discusses the findings; and Section 5 concludes the study.

2. Theoretical Framework and the Literature

2.1. Theoretical Framework

Theoretically, financial risk can influence CO₂ through multiple channels, including investment flows, industrial activities, and policy prioritization. Higher financial risk may dampen economic activities and reduce industrial production, leading to lower energy consumption and emissions in the short run [20,21]. However, persistent financial instability can also divert attention and resources away from climate initiatives and delay investments in clean energy infrastructure, thereby hindering decarbonization efforts and potentially increasing CO₂ in the long run [22]. Thus, the relationship between FR and emissions is theoretically ambiguous, depending on the severity, duration, and policy responses during periods of financial instability. Globalization plays a multifaceted role in shaping environmental outcomes, as it can drive economic growth and industrialization while simultaneously facilitating the transfer of cleaner technologies and environmental standards. Theoretically, globalization may increase CO₂ through the scale effect, as greater trade openness and foreign direct investment boost energy-intensive production and consumption [23,24]. Conversely, globalization can lead to emissions reductions through the technique effect, where technological diffusion and access to environmentally friendly practices reduce the carbon intensity of production [25]. The net effect of globalization on CO₂ thus depends on the interplay between these competing effects and the structural characteristics of the host economy.

Human capital plays a crucial role in shaping CO₂ emissions by influencing the adoption of cleaner technologies, enhancing energy efficiency, and supporting green innovation [26,27]. Higher levels of education and skill development enable individuals and firms to utilize advanced production processes that can reduce reliance on carbon-intensive activities, thereby mitigating emissions [18,28]. However, in the early stages of human capital expansion, rising incomes and industrial activities may initially lead to increased energy demand and higher CO₂ emissions, reflecting a transitional phase before environmental benefits materialize [29]. Thus, the impact of human capital on CO₂ emissions may exhibit a dynamic relationship, contingent on a country’s stage of economic development and the effectiveness of environmental policies.

Renewable energy consumption is widely theorized to reduce CO₂ by substituting fossil fuels with cleaner energy sources, thereby lowering the carbon intensity of the energy mix [30,31]. The adoption of renewable energy technologies supports the transition toward low-carbon economies by providing sustainable energy alternatives, improving energy security, and mitigating environmental degradation. However, in contexts where renewable integration is incomplete or where fossil fuels are used to support renewable infrastructure, the anticipated emissions reduction may be limited or delayed [32,33]. Therefore, while the theoretical expectation is that REEC lowers emissions, the effectiveness of this pathway relies on the degree of renewable penetration and the supporting policy and technological environment within the economy. Fossil fuel efficiency (FFE) relates to the amount of output or energy services derived from each unit of fossil fuel consumed. Theoretically, improving FFE should reduce CO₂ by decreasing the quantity of fossil fuels required for economic activities, thus reducing carbon emissions per unit of output [8,34]. However, the rebound effect suggests that improvements in efficiency can lower the cost of energy services, which may lead to increased fossil fuel consumption and offset the potential emissions reduction [6,35]. In the context of China and other rapidly growing economies, where energy demand is high and industrialization continues, the effectiveness of FFE in reducing emissions may be limited unless accompanied by complementary measures such as carbon pricing and stringent emission standards. This framework underscores the importance of examining the role of FFE within a broader policy context when assessing its impact on CO₂.

2.2. Literature Review

The literature on the nexus between renewable energy consumption (REEC) and CO₂ emissions strongly supports the argument that expanding renewables contributes to environmental sustainability by reducing emissions across different contexts and methodologies. Studies such as [36] for G7 countries, ref. [30] for the U.S. and China, and [37] for developed economies using ARDL, DARDL, and NARDL frameworks consistently report that increased REEC is associated with reduced CO₂ emissions. Similar results from [31] using CUP-FM and [38] employing Tapio decoupling further reinforce the robustness of this relationship across diverse methodological approaches and regions. However, ref. [39], using the QQR method in China, presents a contrasting finding where REEC is associated with increased CO₂ emissions, potentially highlighting transitional inefficiencies and rebound effects in emerging economies. This divergence underscores the importance of regional contexts and the stage of renewable energy integration in determining the effectiveness of REEC in reducing emissions.

Turning to globalization and CO₂ emissions, the literature presents a more nuanced and contested landscape. Studies such as [14,40] argue that globalization tends to exacerbate CO₂ emissions in G7 and Asian countries, driven by increased industrial activity, trade expansion, and fossil fuel consumption associated with deeper global integration. This is consistent with the pollution haven hypothesis, where increased global trade may shift pollution-intensive industries to regions with laxer environmental standards. Conversely, evidence from [41] on BRICS nations and [13] on a broader set of 38 countries shows that globalization can reduce CO₂, possibly due to technology transfer, cleaner production practices, and heightened environmental awareness facilitated by global cooperation. Ref. [42] also reveals mixed findings in top tourist nations, indicating that the effect of globalization on emissions is highly context-dependent, shaped by domestic policies and the nature of integration into global markets.

The impact of financial risk on CO₂ is similarly contested in the literature, reflecting the complex interplay between financial market stability and environmental outcomes. Studies like Zhao et al. (2021) [21], Yang et al. (2020) [43], and Wang et al. (2022) [38] provide evidence that higher financial risk is associated with lower CO₂ emissions, suggesting that economic slowdowns induced by financial instability may reduce energy consumption and, consequently, emissions. This aligns with the argument that financial crises may temporarily suppress industrial output, thereby reducing environmental pressure. However, refs. [33,44] present the opposite view, demonstrating that financial risk can increase CO₂ emissions by undermining environmental governance and delaying green investments. These mixed findings highlight the dual role of financial risk, which may reduce emissions in the short term due to reduced economic activities but may also hinder long-term environmental improvements by diverting resources away from sustainable development projects.

In the context of energy efficiency (EF) and CO₂, the literature is notably consistent, reinforcing the argument that improvements in energy efficiency contribute to emission reductions. Ref. [45] in Sweden, ref. [7] in developed nations, ref. [34] across 120 countries, and [46] in hydrogen-consuming countries all report that enhanced energy efficiency leads to lower CO₂ emissions, regardless of the methodological approaches employed, including wavelet tools, Fourier ARDL, QR, and AMG. These consistent findings underscore the potential of energy efficiency as a crucial tool for achieving decarbonization goals, supporting the argument for its prioritization within climate policies to reduce carbon footprints while maintaining economic productivity.

Table 1. Summary of studies.

Author(s)	Periods	Nation(s)	Method(s)	Finding(s)
CO2 and REEC
[36]	1979–2019	G7 countries	ARDL	REEC ↓ CO2
[30]	1980–2016	U.S. and China	DARDL	REEC ↓ CO2
[37]	1990–2018	22 well-developed countries	NARDL	REEC ↓ CO2
[31]	1995–2018	E∓7 economies	CUP-FM	REEC ↓ CO2
[38]	2005–2019	BRI nations	Tapio decoupling	REEC ↓ CO2
[47]	1980–2019	MENA nations	IRF	REEC ↓ CO2
[32]	1997–2017	U.S.	MMQR	REEC ↓ CO2
[39]	1980–2017	China	QQR	REEC ↑ CO2
CO2 and GLOB
[14]	1995–2019	G7 countries	ARDL	GLOB ↑ CO2
[37]	1960–2020	Asian countries	PCSE	GLOB ↑ CO2
[41]	1991–2022	BRICS countries	Panel Quantile Regression	GLOB ↓ CO2
[42]	2000–2020	Top tourist nations	Panel Regression	GLOB ↓↑ CO2
[12]	1990–2019	Somalia	ARDL	GLOB ↑ CO2
[13]	1992–2019	38 countries	DCCE-MG	GLOB ↓ CO2
CO2 and FR
[21]	2003–2018	62 countries	Panel Quantile Regression	FR ↓ CO2
[43]	1980–2016	54 developing economies	GMM	FR ↓ CO2
[33]	2020–2023.	Globe	WLMC	FR ↑ CO2
[44]	2000–2023	87 countries	Panel Analysis	FR ↑ CO2
[38]	1984–2020	China	ARDL	FR ↓ CO2
CO2 and EF
[45]	1990–2020	Sweden	Wavelet tools	EF ↓ CO2
[7]	1990–2020	Developed nations	Fourier ARDL	EF ↓ CO2
[34]	2010–2020	120 countries	QR	EF ↓ CO2
[46]	1995–2019	Hydrogen-consuming nations	AMG	EF ↓ CO2

presents a summary of the findings.

2.3. Evaluation of the Literature

After a careful review of the literature, the following was observed: First, under methodological contrasts, ARDL, DARDL, NARDL, CUP-FM, and Tapio decoupling studies (e.g., [30,31,36,37,38]) uniformly find that renewable energy consumption reduces CO₂—yet China’s QQR result [39] bucks this trend, signaling transitional inefficiencies inherent to its rapid renewables rollout. Second, in the globalization debate, ARDL and PCSE analyses of G7 and Asian countries [14,37] show globalization raising emissions, whereas Panel Quantile Regression in BRICS [41] and DCCE-MG over 38 countries [13] observe the opposite; clustering by export composition would explain why China’s pollution-intensive trade profile aligns more with the former. Third, regarding financial risk, panel quantile and GMM studies across 62 and 54 countries [21,43] link higher risk to lower CO₂—mirroring China’s ARDL evidence [38]—while WLMC and Panel Analysis of the globe [33,44] indicate that risk can also elevate emissions by stalling green investment. Finally, consistent EF–CO₂ findings across Sweden, developed nations, 120 countries, and hydrogen-consuming states [7,34,45,46] underscore energy efficiency’s universal decarbonization role, which in China must be examined against its mix of centrally planned and market-driven provinces. This thematic, China-focused framing highlights where domestic policy and industrial constraints most shape each nexus. Based on these studies, no research has yet focused exclusively on China using cross-quantile analysis for both dependent and independent variables. Likewise, the environmental impact of fossil fuel efficiency in the Chinese context remains unexplored, leaving a gap that this study aims to fill.

Based on the four guiding questions, we propose the following testable hypotheses:

 H1 (Economic Growth):

Higher rates of economic growth in China are associated with increases in CO₂ emissions.

 H2 (Globalization):

Greater degrees of economic globalization (trade and financial integration) exacerbate China’s CO₂ emission trends.

 H3 (Fossil Fuel Efficiency):

Improvements in fossil fuel efficiency (FFE) lead to reductions in China’s CO₂ emissions.

 H4 (Human Capital):

Higher levels of human capital (e.g., educational attainment, skill indices) are associated with lower CO₂ emissions in China.

 H5 (Renewable Energy):

Increased renewable energy consumption contributes to a decrease in China’s CO₂ emissions.

3. Data and Methods

3.1. Data

This study investigates the key determinants of CO₂ emissions in China over the period 1984Q1 to 2023Q4. The selected explanatory variables include economic growth (EG), financial risk (FR), globalization (GLOB), human capital (HC), renewable energy consumption (REEC), and fossil fuel efficiency (FFE). The definitions, measurement approaches, and data sources for each variable are detailed in

Table 2. Descriptive statistics.

	Full Name	Measurement	Sources
CO2	CO2 emissions	Per Capita	[48]
EG	Economic growth	GDP Per Capita USD 2015	[49]
FR	Financial risk	Index	[50]
GLOB	Globalization	Index	[51]
REEC	Renewable energy consumption	Per Capita (Kwh)	[48]
FFE	Fossi fuel efficiency	${\displaystyle \frac{\mathrm{G}\mathrm{D}\mathrm{P} \mathrm{P}\mathrm{e}\mathrm{r} \mathrm{C}\mathrm{a}\mathrm{p}\mathrm{i}\mathrm{t}\mathrm{a}}{\mathrm{F}\mathrm{o}\mathrm{s}\mathrm{s}\mathrm{i}\mathrm{l} \mathrm{F}\mathrm{u}\mathrm{e}\mathrm{l} \mathrm{C}\mathrm{o}\mathrm{n}\mathrm{s}\mathrm{u}\mathrm{m}\mathrm{p}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n} \mathrm{P}\mathrm{e}\mathrm{r} \mathrm{C}\mathrm{a}\mathrm{p}\mathrm{i}\mathrm{t}\mathrm{a}}}$	Author’s Calculation
HC	Human capital	Index	Penn Database

. The sample period is determined by data availability, with FR data beginning in 1984 and a lack of consistent data for all variables beyond 2023Q1. China is selected due to its position as the world’s largest emitter of CO₂ and its significant role in global energy consumption and economic growth. Its ongoing transition toward renewable energy and policy reforms offers a compelling context to examine the drivers of emissions.

Figure 1 provides a comprehensive overview of the descriptive statistical characteristics of CO₂ emissions and their key determinants—EG, FR, GLOB, REEC, FFE, and HC—in China from 1984Q1 to 2023Q4. In terms of central tendency, EG exhibits the highest mean (8.00) and median (7.98), reflecting robust economic growth, while REEC also shows a high mean (6.84) and median (6.57), indicating significant renewable energy adoption. CO₂ presents lower central values (mean: 1.39; median: 1.36), and FFE maintains the lowest mean (0.26), suggesting limited fossil fuel efficiency improvements over the period. EG records the highest maximum value (9.43), while FFE shows the lowest minimum (0.11). Regarding variability, REEC has the highest standard deviation (0.99), indicating substantial fluctuations in renewable energy consumption, followed closely by EG (0.94). Distributionally, the Jarque–Bera, skewness, and kurtosis statistics reveal notable deviations from normality across several variables. FR, in particular, demonstrates extreme non-normality with a high negative skewness (−2.72), high kurtosis (14.63), and an exceptionally large Jarque–Bera statistic (1099.28). By contrast, CO₂ and EG exhibit near-symmetric distributions, with skewness values close to zero and moderate kurtosis, suggesting more stable and consistent behavior across the study period.

3.2. Empirical Method

3.2.1. Modified Cross-Quantile Regression

The study utilizes the recently proposed Modified Cross-Quantile Regression (MCQR) by [19], which advances the Cross-Quantile Regression (QQR) framework originally developed by [52]. While QQR has been widely applied, it is constrained by bandwidth selection issues, as highlighted by Sim and Zhou (2015) [52]. Specifically, QQR estimates are highly sensitive to the choice of bandwidth, and the commonly used bandwidth of 0.05 can lead to singular design matrix errors. For instance, although a 0.05 bandwidth may yield estimable QQR results when using log-first-difference series, it often results in singularity errors when applied to percentage log-first-difference series and produces identical results when the bandwidth is increased to 5. The MCQR approach addresses these limitations, offering a more robust and reliable framework for assessing the relationship between CO₂ emissions and their determinants.

Beyond the bandwidth sensitivity limitation, the QQR framework lacks a formal structure for assessing the statistical significance of its parameter estimates. Although studies such as [53] report significant and insignificant QQR outcomes, they do not resolve the underlying sensitivity of QQR results to bandwidth choices, which may compromise the robustness of the estimates. To address these concerns, ref. [54] introduced the MCQR approach, which mitigates both bandwidth dependency and inference limitations inherent in the standard QQR method. The CQR technique begins by constructing the Quantile Series of Y and the X using the Quantile Series (QSER) method proposed by [55]. Subsequently, it implements regression across all quantile combinations of X and Y, enabling the analysis to uncover the complete distributional dependence between the variables. Formally, the MCQR is modeled as follows:

$$Q_{\tau }\left(Y\right)={\mathsf{{\rm Y}}}_0(\tau ,\theta )+{\mathsf{{\rm Y}}}_1(\tau ,\theta )Q_{\theta }\left(X\right)+ϵ(\tau ,\theta )$$

(1)

where the conditional $\mathsf{\tau }$-th quantile of $Y$ is represented by ${\mathrm{Q}}_{\mathsf{\tau }}\left(Y\right)$. The conditional $\mathsf{\theta }$-th quantile of $X$ is denoted by ${\mathrm{Q}}_{\mathsf{\theta }}\left(X\right)$. The intercept is denoted by ${{\rm Y}}_0\left(\mathsf{\tau },\mathsf{\theta }\right)$, while slope coefficient is represented by ${{\rm Y}}_1\left(\mathsf{\tau },\mathsf{\theta }\right)$, and $ϵ\left(\mathsf{\tau },\mathsf{\theta }\right)$ signifies the error term.

The MCQR framework improves on traditional Quantile Regression by capturing dependence across the full distributions of both variables, revealing how shocks at specific quantiles of the independent variable affect different quantiles of the dependent variable. This helps identify tail behaviors and asymmetries that standard methods may miss. To check robustness, this study compares the average CQR estimates with MQR estimates.

3.2.2. Quantile-on-Quantile KRLS

A convenient way to frame Quantile-on-Quantile KRLS is via its penalized loss in an $\mathcal{H}\mathrm{with}\mathrm{kernel}K.\mathrm{Let}{Q}_{\nu }\left(X_i\right)$ denote the $\nu \text{-}\mathrm{th}\mathrm{quantile}\mathrm{of}\mathrm{the}\mathrm{covariate}{X}_i$. Then, the estimator $f_{\tau ,\nu }\in \mathcal{H}$ solves

$${\widehat{f}}_{\tau ,\nu }=\mathrm{a}\mathrm{r}\mathrm{g} \underset{f\in \mathcal{H}}{min} \sum _{i=1}^n {\rho }_{\tau }\left(Y_i-f\left(Q_{\nu }\left(X_i\right)\right)\right)+\lambda \Vert f{\Vert }_{\mathcal{H}}^2$$

(2)

where

$${\rho }_{\tau }\left(u\right)=u(\tau -1\{u<0\left\}\right)$$

(3)

is the quantile (check) loss and $\lambda$ is the regularization parameter. By the representer theorem, the solution admits the expansion

$${\widehat{f}}_{\tau ,\nu }(x)=\sum _{i=1}^n {\alpha }_{i}K\left(Q_{\nu }\left(X_i\right),x\right),$$

(4)

so that the estimated τ-th conditional quantile of Y given the ν-th quantile of X is

$${\widehat{Q}}_{\tau }\left(Y\mid Q_{\nu }\left(X\right)=x\right)=\sum _{i=1}^n {\alpha }_{i}K\left(Q_{\nu }\left(X_i\right),x\right).$$

(5)

Figure 2 presents the flow of the analysis.

4. Findings and Discussion

4.1. Forecast Results

The study begins by evaluating the predictive power of the regressors on CO₂. To achieve this, we employed Linear Regression, Random Forest, and Support Vector Regression models (see Figure 3). In Figure 3a, which explores the relationship between CO₂ and REEC, both RF and SVR perform exceptionally well with low RMSE values (0.06) and high R² scores (0.99), outperforming the Linear model (RMSE: 0.12, R²: 0.95). Figure 3b presents a moderate drop in prediction accuracy when using FFE, with RF and SVR again performing better (R² = 0.85 and 0.90) compared to Linear (R² = 0.83). Figure 3c shows strong predictive power for all models with GLOB, with RF and SVR achieving RMSE = 0.08 and R² = 0.98, significantly better than Linear (RMSE = 0.22, R² = 0.84). However, Figure 3d reveals a noticeable performance gap when using FR as the predictor. The Linear model performs poorly (R² = 0.48), while RF and SVR yield better results (R² = 0.80 and 0.72, respectively), though their RMSEs are higher (0.25 and 0.29). Figure 3e shows that EG is the most effective regressor, with all models achieving near-perfect prediction: RF (RMSE: 0.02, R²: 0.99), SVR (RMSE: 0.04, R²: 0.99), and Linear (RMSE: 0.09, R²: 0.97). Figure 3f shows that HC is an effective regressor, with all models achieving high predictive accuracy: Linear (RMSE: 0.16, R²: 0.90), RF (RMSE: 0.03, R²: 1.00), and SVR (RMSE: 0.08, R²: 0.98). Overall, Random Forest and SVR consistently outperform Linear Regression across all variables in terms of lower RMSE and higher R², highlighting the strength of nonlinear models in capturing complex relationships between CO₂ emissions and their determinants.

4.2. Results of Nonlinearity and Normality Tests

Table 3. BDS statistics.

	CO2	EG	FR	FFE	GLOB	REEC	HC
M2	52.096 ***	41.709 ***	11.386 ***	44.237 ***	43.397 ***	29.614 ***	53.233 ***
M3	55.301 ***	44.237 ***	13.524 ***	46.897 ***	46.487 ***	31.204 ***	56.773 ***
M4	59.516 ***	47.661 ***	15.270 ***	50.449 ***	50.350 ***	33.417 ***	61.821 ***
M5	65.823 ***	52.838 ***	16.734 ***	55.852 ***	55.925 ***	36.853 ***	69.098 ***
M6	74.573 ***	60.062 ***	18.589 ***	63.376 ***	63.547 ***	41.686 ***	79.032 ***

Note: *** p < 0.01.

presents the BDS (Brock–Dechert–Scheinkman) test results for CO₂ and its regressors—EG, FR, FFE, GLOB, HC, and REEC—across embedding dimensions M2 to M6. The consistently high and statistically significant test statistics for all variables across all dimensions indicate strong evidence of nonlinearity in the underlying data-generating processes. The null hypothesis of the BDS test states that the series is independently and identically distributed (i.i.d.). Given the significance levels reported, the null hypothesis is rejected for all variables, confirming that the dynamics of CO₂ emissions and their explanatory variables are nonlinear and not purely random.

The study also used the Q-Q plots (see Figure 4) to assess the normality of each variable’s distribution by comparing sample quantiles to theoretical quantiles from a normal distribution. In all plots, the data points deviate significantly from the reference 45-degree line, particularly at the tails, indicating clear departures from normality. The null hypothesis of the Q-Q plot analysis is that the data are normally distributed. Based on the observed nonlinearity and curvature in each plot, the null hypothesis is rejected for all variables, confirming that none of the series follows a normal distribution. This supports the need for non-parametric or robust modeling approaches.

4.3. Stationarity Test Result

Next, we tested the stationarity attributes of the series using the quantile Phillips–Perron (PP) test (see Figure 5) to evaluate the stationarity of CO₂ and its regressors (EG, FR, FFE, GLOB, REEC) across different quantiles (0.05 to 0.95). The null hypothesis of the PP test states that the series is non-stationary (i.e., has a unit root), while the alternative hypothesis is that the series is stationary. In the plot, the critical values at 1%, 5%, and 10% significance levels are represented by the shaded grey area. Test statistics that fall below these thresholds indicate rejection of the null hypothesis. Based on the graph, most variables are non-stationary in the lower quantiles but tend to become stationary at higher quantiles—particularly EG and REEC, which show consistent rejection of the null in upper quantiles. This suggests that the stationarity properties of the variables are quantile-dependent, affirming the usefulness of quantile-based approaches for further analysis.

4.4. Cross-Quantile Regression Result

After confirming the nonlinear characteristics of the series, we employed the recently developed Cross-Quantile Regression (CQR) method to examine the relationship between CO₂ emissions and their regressors—REEC, FR, EG, FFE, and GLOB. Figure 6 presents the results of CQR.

Figure 6a presents the effect of REEC on CO₂ emissions across their respective quantiles in China. The result shows consistently positive and statistically significant association at all quantile levels. This suggests that higher levels of REEC are associated with higher CO₂, particularly when both REEC and CO₂ are in the upper quantiles. This counterintuitive positive relationship may reflect structural and transitional issues in China’s energy sector, where renewable energy expansion is occurring alongside continued reliance on fossil fuels or carbon-intensive infrastructure needed to support renewables (e.g., coal backup for intermittent solar or wind). Such findings are not unprecedented. Studies like [39,56] have observed similar trends, attributing them to the early-stage integration of renewables in developing and transitional economies. In China’s case, large-scale renewable deployment often still coexists with high baseline emissions from coal-dominated grids and industrial demand, which can dilute the net effect of renewables. Additionally, the manufacturing and deployment of renewable infrastructure itself (e.g., solar panel production) can initially contribute to emissions. Conversely, studies like [57] argue that over time, as grid efficiency improves and fossil fuel subsidies decline, this relationship may eventually turn negative. Hence, the observed positive association could be reflective of a transitional phase where renewable expansion has not yet been accompanied by proportional reductions in carbon-intensive energy sources.

Figure 6b showcases the impact of fossil fuel efficiency (FFE) on CO₂ across quantiles of both variables in China. In general, we observed a strong, positive, and statistically significant relationship across most quantile combinations—particularly in the middle-to-upper quantiles of both variables. This indicates that as fossil fuel efficiency improves (i.e., more energy is extracted per unit of fossil fuel), CO₂ also tends to increase, especially at higher emission levels. This seemingly paradoxical finding implies that gains in FFE may not necessarily result in lower emissions in China’s current energy context. This positive association can be explained by the rebound effect or Jevons paradox, where efficiency improvements reduce the cost of energy services and consequently increase fossil fuel usage, offsetting potential emission reductions [8]. In China, fossil fuel efficiency improvements may actually enable greater industrial output and higher fossil fuel consumption overall, especially when not accompanied by strict emission limits or clean energy transitions. Studies like [5] support this view, showing that energy efficiency improvements in China’s coal and oil sectors often lead to increased CO₂ due to accelerated production and consumption. On the other hand, critics such as [6] argue that over the long term, fossil fuel efficiency contributes to decoupling emissions from growth, particularly when combined with regulatory interventions. However, the consistently strong and positive coefficients suggest that in the short-to-medium term, FFE in China may inadvertently exacerbate CO₂, particularly when efficiency gains are used to scale up fossil-fuel-driven output.

Figure 6c displays the impact of globalization (GLOB) on CO₂ across quantiles of both variables in China. We observed that the relationship between GLOB and CO₂ is consistently positive and statistically significant across nearly all quantiles. Furthermore, the intensity of the association deepens along higher quantiles of GLOB, implying that the more China integrates into the global economy, the stronger the increase in CO₂, particularly when emissions are already high. This positive relationship may be attributed to the scale effect of globalization in China, where deeper integration into global trade and investment flows has led to a substantial rise in industrial production, energy consumption, and thus emissions. China’s role as a global manufacturing hub—powered primarily by fossil fuels—amplifies this environmental burden [58]. This finding aligns with [59,60,61], where it was found that globalization intensifies environmental degradation in emerging economies unless accompanied by clean technologies and environmental regulations. However, this contrasts with studies like [24], which suggests that globalization may also foster the diffusion of green technologies and environmentally sustainable practices in the long run. In the case of China, the current trend shown in the heatmap suggests that the environmental cost of globalization still outweighs its technological benefits, especially when global economic engagement increases without parallel environmental governance.

Figure 6d unveils the interrelationship between financial risk (FR) and CO₂ across various quantiles of both variables in China. The result reveals a positive and significant relationship across nearly all quantile combinations, especially when FR is in the upper quantiles. Notably, darker shades in the upper-left region of the plot suggest that higher levels of financial risk are associated with larger increases in CO₂, particularly when emissions are relatively low to moderate. This positive association may be attributed to the destabilizing effect that financial uncertainty imposes on environmental governance and sustainable investments. In China, elevated financial risk—driven by factors like debt accumulation, credit market volatility, or policy uncertainty—can undermine green investment initiatives and shift focus toward short-term economic stabilization, often relying on carbon-intensive industries [60]. When capital markets are volatile or constrained, firms may delay or abandon clean energy projects in favor of cheaper, emission-heavy alternatives. This aligns with findings by [20,22], who observed that financial instability tends to weaken environmental performance by restricting the flow of capital to sustainable sectors. However, opposing views, such as [22], suggest that heightened financial risk could eventually compel regulatory tightening and more efficient resource use, potentially reducing emissions.

Figure 6e highlights the relationship between economic growth (EG) and CO₂ across quantiles of both variables in China. The outcomes reveal a consistently positive and significant relationship between EG and CO₂ emissions across all quantile interactions, though the magnitude varies slightly. This suggests that as economic growth increases, CO₂ emissions also rise, regardless of whether the economy or emissions are at low, medium, or high levels. This positive association reflects China’s growth model, which has historically relied heavily on energy-intensive manufacturing, export-oriented industrialization, and fossil fuel consumption. Rapid urbanization and infrastructure expansion have led to increased energy demand, contributing to higher CO₂ emissions [62]. The finding is consistent with the Environmental Kuznets Curve (EKC) theory in its early stages, where economic growth initially leads to environmental degradation. Studies such as [63,64,65] support this result, noting that in developing and emerging economies like China, economic expansion is often carbon-intensive. However, others argue that with structural transformation, technological upgrades, and a shift toward services and renewables, this relationship could weaken or even reverse over time [66]. Nonetheless, the result demonstrates that at present, economic growth in China continues to exert upward pressure on CO₂ emissions, highlighting the need for green growth policies and decarbonization strategies.

Figure 6f presents the effect of human capital on CO2. The result indicating a significant positive connection across the quantiles suggests that human capital, rather than mitigating environmental degradation, is contributing to increased CO₂ emissions in China. This implies that as education levels and workforce skills improve, there is a corresponding rise in economic activities such as industrial output and energy consumption, which, in China’s current fossil-fuel-heavy energy structure, translates into higher emissions. The quantile-based consistency of this result further indicates that this relationship holds not only at average emission levels but also in both lower and higher emission scenarios, emphasizing the systemic nature of this interaction in China’s growth dynamics. These findings align with the study by [67], which found that in China, higher human capital contributes to economic expansion that heavily relies on energy-intensive and emission-intensive sectors, thereby increasing CO₂ emissions. Similarly, ref. [26] highlighted that while human capital is essential for technological advancement, without a parallel transition toward cleaner energy, its contribution to economic growth can exacerbate environmental degradation in developing economies. Thus, the observed positive link underscores the need for China to align human capital development with green technology adoption and clean energy policies to decouple growth from emissions effectively.

Table 4. Summary of findings.

Figure	Variable	Relationship with CO2	Significance	Explanation and Interpretation
Figure 6a	Renewable Energy Expansion	Positive	Significant across all quantiles	Transitional phase where renewable energy expansion coexists with carbon-intensive infrastructure, leading to increased emissions despite renewable deployment.
Figure 6b	Fossil Fuel Efficiency	Positive	Significant, stronger in mid-to-upper quantiles	Reflects Jevons paradox; efficiency gains lower energy costs, increasing fossil fuel use and emissions.
Figure 6c	Globalization	Positive	Significant, stronger in higher quantiles	Driven by scale effect of China’s global trade integration, causing higher industrial production and emissions.
Figure 6d	Financial Risk	Positive	Significant, strongest in upper quantiles of FR	High financial risk undermines sustainable investments, increasing short-term reliance on carbon-intensive industries.
Figure 6e	Economic Growth	Positive	Significant across all quantiles	Energy-intensive growth model consistent with early Environmental Kuznets Curve stage, highlighting the need for sustainable growth policies.
Figure 6f	Human Capital	Positive	Significant across all quantiles	Improved human capital increases economic activities dependent on fossil fuels, thereby intensifying emissions without a green energy shift.

presents a summary of the findings from the CQR.

4.5. Robustness Check with MQR and ACQR

Next, we compare the results of Modified Quantile Regression (MQR) and average Cross-Quantile Regression (QQR, blue) between CO₂ emissions and their key determinants in China (see Figure 7) from 0.05 to 0.95. In Figure 7a,b,e,f, the MQR estimates (red) for CO₂ and REEC, FFE, EG, and HC show a downward trend across quantiles, indicating that the positive influence of REEC, FFE, and EG on CO₂ emissions decreases in higher emission quantiles. For instance, the slope coefficients for FFE in Figure 7b are notably higher across all quantiles (ranging from ~4.5 in lower quantiles to ~2 in higher quantiles), reflecting its stronger positive association with CO₂ emissions in lower quantiles. The ACQR (blue) estimates remain consistently lower across the quantiles, highlighting that MQR captures stronger and more dynamic effects, particularly in lower quantiles. In Figure 7c,d, representing CO₂ with GLOB and FR, respectively, the MQR estimates also decline across quantiles, with GLOB showing a steeper decline, suggesting that globalization’s contribution to CO₂ emissions is more pronounced in lower quantiles and diminishes as emissions increase. For FR, the MQR estimates remain consistently low (~1.3 to ~0.5), indicating a relatively weak but negative association with CO₂ across the quantiles, while the ACQR estimates are consistently higher, indicating that the average quantile approach may overstate the association compared to the conditional quantile-specific estimates. Overall, the divergence between MQR and ACQR estimates emphasizes the importance of distributional heterogeneity in the CO₂ emissions interrelationship across quantiles, with MQR capturing nuanced dynamics missed by standard Quantile Regression averages, indicating that policy interventions may need to be tailored according to emission levels to effectively manage CO₂ emissions in China.

4.6. Quantile-on-Quantile KRLS Results

Next, we employed the Quantile-on-Quantile Kernel Regularized Least Squares (QQ-KRLS) method (see Figure 8) to examine the association between CO₂ and its determinants. Starting from Figure 8a, the connection between CO₂ and REEC shows a positive relationship across the distribution. The deeper green in the lower quantiles of CO₂ suggests that REEC has a stronger positive effect when emissions are low, potentially reflecting the transitional phase where renewable adoption coexists with fossil fuel dependence in China. In Figure 8b, which examines CO₂ and FFE, a more complex pattern emerges. The middle quantiles of both CO₂ and FFE exhibit strong positive effects, indicating that fossil fuel efficiency is positively associated with CO₂ under moderate emission and efficiency conditions—a phenomenon often tied to the rebound effect where efficiency gains lead to increased energy use. In contrast, the lower and upper quantiles of FFE display red shading, signaling negative associations with CO₂ under extreme efficiency levels. This indicates that at very high or low levels of fossil fuel efficiency, the relationship with emissions may invert, likely due to structural shifts in energy use or technological thresholds that reduce emissions despite efficiency changes. Moving to Figure 8c, the relationship between CO₂ and GLOB shows a general positive association but with areas of red indicating negative impacts around the mid–high quantiles of CO₂ and high quantiles of GLOB, suggesting globalization may reduce emissions under certain high-emission scenarios, possibly through technology transfer or green investment channels. In Figure 8d with FR, positive impacts are clustered in the high quantiles of both FR and CO₂, highlighting that financial risk significantly influences emissions during high-risk, high-emission periods, while lower quantiles reflect weaker and less consistent effects. Figure 8e on CO₂ and EG and Figure 8f on CO₂ and HC similarly displays a persistent positive relationship across nearly all quantiles, with a stronger effect in lower quantiles of CO₂, indicating that economic growth and human capital contribute more to emissions when emission levels are low, while the effect moderates as emissions increase.

5. Conclusions and Policy Recommendations

5.1. Conclusions

Fossil fuel efficiency, human capital, renewable energy adoption, and globalization are reshaping the carbon landscape in complex and sometimes counterintuitive ways—yielding emissions reductions in some segments while creating latent trade-offs in others—and ultimately determining whether an economy can decouple growth from CO₂ emissions. This study examines these drivers of CO₂ emissions in China over 1984 Q1–2023 Q4. We employ the recently proposed Modified Cross-Quantile Regression by [19] to capture heterogeneous dependence structures across the distributions of both independent and dependent variables, and we validate our results with Quantile-on-Quantile KRLS and Modified Quantile Regression as robustness checks. To our knowledge, this is the first empirical analysis investigating how fossil fuel efficiency, human capital, renewable energy consumption, and globalization jointly influence CO₂ emissions in China. Our findings reveal that globalization, fossil fuel efficiency gains, renewable energy expansion, and financial risk all exhibit a positive association with CO₂ emissions across various distributional segments.

5.2. Policy Recommendations

In the power-generation sector, particularly in wind-rich provinces like Gansu and Inner Mongolia, policymakers can immediately mandate coal-fired plants to limit their ramp-up for grid stability while incentivizing battery-storage installations and demand-response pilots in cities such as Shanghai under the “Internet + Smart Grid” initiative. Short-term actions include expanding the State Grid’s pilot energy-storage tenders and adjusting transmission tariffs to favor renewables. Over the longer term, systemic reform of coal-backup norms—phasing out must-run coal contracts—and adoption of European-style ancillary services markets must be tailored to China’s regulated pricing regime and provincial budget constraints. While Germany’s coal phase-out blueprint offers insights, China must adapt it to its market-state hybrid model by strengthening provincial renewable-integration targets rather than imposing a uniform national timeline.

In energy-intensive industries (e.g., steel in Hebei, cement in Shanxi), strict emissions-cap pilot zones—building on the existing seven provincial carbon trading schemes—can translate efficiency gains into absolute reductions. Short-term, local regulators can impose declining emissions allowances tied to energy efficiency benchmarks (e.g., top-quartile furnace performance) while applying incremental coal-tax surcharges. For enduring impact, China must integrate a national carbon-border adjustment mechanism calibrated to its export mix, rather than copying the EU’s complex tariff system outright. Institutional constraints—such as fragmented provincial compliance enforcement—require capacity-building in local environmental bureaus before national cap-and-trade expansion.

China’s export-oriented manufacturing clusters in Guangdong and Zhejiang can benefit immediately from “clean goods” preferential financing under existing Export Credit Insurance Corporation pilots for green tech exports. In the medium term, negotiators should embed technology-transfer clauses in Regional Comprehensive Economic Partnership agreements to secure licensing of advanced photovoltaic and battery technologies, as demonstrated by China–Australia renewable MoUs. Adopting a U.S. carbon-border adjustment model wholesale risks retaliation; instead, a bespoke carbon intensity labelling scheme for domestic exporters—aligned with China’s Made in China 2025 clean tech targets—can drive decarbonization without provoking trade disputes.

To safeguard renewable energy investment during market stress, provincial governments (e.g., in Jiangsu and Fujian) can immediately deploy state-backed green bonds under the national Green Bond Endorsed Project Catalogue, coupled with mandatory climate-risk disclosures for SOEs. Over the long term, China should evolve its green-finance framework into a countercyclical tool—linking PBOC lending reserve requirements to green credit growth—rather than importing the EU’s taxonomy in rigid form. Addressing institutional gaps in the risk assessment capabilities of rural commercial banks is essential before scaling green credit guarantees nationwide.

In heavy-industry provinces like Shanxi and Liaoning, short-term measures include reallocating provincial GDP performance scores to favor high-value services and clean tech manufacturing under the “Dual Circulation” strategy. In the medium term, central planners must pilot sectoral carbon budgets—above and beyond the national CO₂ peak target—in key regions, integrating these into Five-Year Plan targets. Long-run structural transformation requires reforming local land-use and infrastructure subsidies that currently favor coal plants; global green growth models (e.g., Denmark’s wind industrial cluster) provide useful principles but must be adapted to China’s provincial fiscal system and land-use rights framework.

5.3. Limitations and Future Recommendations

A key limitation of this study is that while the quantile-based analysis captures distributional heterogeneity, it does not account for time-varying associations in China. The cross-quantile framework used provides valuable insights into the heterogeneity of relationships across quantiles but remains static, overlooking how these associations may evolve in response to technological advancements, policy interventions, or structural shifts in the economy over time. This limitation restricts the ability to assess dynamic impacts and transition phases in China’s decarbonization efforts. Additionally, the study does not disaggregate emissions by sector, limiting the precision in identifying industry-specific drivers within quantile levels. Data constraints, potential measurement errors, and the exclusion of other environmental factors may further affect the generalizability of the findings. Future research could build on this work by employing time-varying quantile methods or rolling-window quantile approaches to investigate how these associations change over time, providing deeper insights for dynamic policy design and adaptive carbon mitigation strategies.

Regarding human capital’s positive effect on CO₂, China should prioritize aligning human capital development with its green transition strategy to mitigate the unintended rise in CO₂ emissions associated with improvements in education and workforce skills. Policymakers should embed environmental sustainability within education and vocational training by integrating green technology, energy efficiency practices, and sustainable industrial processes into curricula and skill development programs. Additionally, incentives should be provided for industries to absorb this skilled workforce into sectors that prioritize clean energy and low-carbon technologies, ensuring that human capital growth translates into green productivity rather than emissions-intensive activities. Strengthening collaborations between universities, technical colleges, and clean energy industries can accelerate research, innovation, and deployment of renewable technologies, enabling China to leverage its human capital gains while advancing toward its carbon neutrality goals and fulfilling its commitments under SDG 13 (Climate Action) and SDG 9 (Industry, Innovation, and Infrastructure).