Introduction

In recent years, the world has been battling potentially catastrophic effects of climate change. The average temperature increased by 0.7 °C1. These environmental changes are the result of rapid and extensive industrialization of high-income and developing countries. This alarming situation has led the United Nations to adopt an agenda to reduce global CO2 emissions as a sustainable development goal (SDGs). China has undergone rapid industrial development in recent decades, which has resulted in environmental degradation. China, the world’s largest emitter of CO₂, is pivotal in the global efforts to combat climate change. The nation contributes over 28% of global emissions owing to its rapid industrialization and dependence on coal-based energy production2.

Notwithstanding its economic expansion, China is pressured to transition to a more sustainable energy model. The dependency on fossil fuels, particularly coal, remains a significant obstacle in achieving substantial emission reductions3,4. However, China committed to reaching a carbon peak by 2030 and carbon neutrality by 20605. These efforts correspond with the United Nations’ Sustainable Development Goals (SDGs), including Goal 13: Climate Action, which advocates immediate actions to address climate change6. Despite these ambitious targets, carbon-neutrality path remains complex as energy infrastructure relies heavily on fossil fuels, with coal accounting for more than half of the energy mix. Furthermore, China continues to provide subsidies for nonrenewable energy sources, hindering the competitiveness of renewable alternatives7.

To address these challenges, coordinated policy interventions are essential. Green innovation can drive technological advancement toward low-carbon solutions, renewable energy adoption can displace fossil fuel consumption, environmental taxation can internalize pollution costs and incentivize cleaner production, and financial development can channel capital toward sustainable investments. However, the effectiveness of these policy tools depends critically on their interaction and implementation context, particularly given China’s vast regional disparities in economic structure, resource endowments, and institutional capacity.

Financial development has gained global acknowledgement as a crucial tool for fostering sustainability and accelerating the shift to a low-carbon economy8. However, the overall effect on CO2 emissions remains a subject of debate. According to Ref9., a healthy financial sector helps reduce emissions by funding sustainable businesses, renewable energy technologies, and energy efficiency programs. Conversely, alternative research indicates that readily available finance and investment prospects may result in heightened industrial activity, overexploitation of resources, and the growth of carbon-intensive industries, ultimately intensifying emissions. Umar et al.10 argued that robust financial systems are crucial for promoting carbon–neutral investments. Financial development reduces carbon emissions through innovation and industrial structures11,12,13,14,15,16. Chen et al.3,4 contended that improving financial frameworks could expedite the transition to low-carbon emissions, highlighting the significance of diverse financial instruments to facilitate adaptation investments. Research by Zaidi et al.17 and Zafar et al.18 indicates that financial development can potentially decrease carbon emissions in both the short and the long run.

However, Al-Mulali et al.19 findings show that financial development can increase emissions in the long term. Zhang and Razzaq20 linked the increase in emissions to the expansion of the financial sector, highlighting the significant role of financial intermediation. Shahbaz et al.21 demonstrated a link between the development of the banking sector and an increase in emissions through unidirectional causality, a trend similar to that observed in Hao et al.22, outcomes across 29 Chinese provinces. Chen et al.23 demonstrated that green bonds, climate finance, and sustainability loans are essential for financing renewable energy initiatives, energy-efficient infrastructure, and technological advancements. The empirical evidence indicates that financial development and environmental regulations influence their efficacy in reducing emissions24. Robust environmental legislation facilitates financial development and validates green initiatives and low-carbon technologies. Without stringent policies, financial growth may merely expedite carbon-intensive economic activity9. In China, financial development reduces carbon emissions through heterogeneous technological progress, with regional differences influencing the impact of financial development on carbon emissions25,26,27 but increasing emissions in certain provinces, with two-way causality in the long term28. Wang et al.29 highlighted the importance of sustainable finance tools in reducing carbon intensity in China’s industrial sector indicating that finance in green industries requires regulatory incentives to ensure sustainability. Conversely, regions that lack green financing initiatives frequently experience financial expansion, which promotes high emission enterprises and exacerbates environmental deterioration. Thus, while financial development has the potential to reduce carbon emissions, its impact depends on policy formation, sectoral resource allocation, and sustainability criteria in financial decision making.

Green innovation has been widely recognized as an effective strategy for mitigating emissions13,14,15. Empirical research has demonstrated that green innovation is a viable strategy for achieving high-quality economic development and safeguarding the environment30. Zhai et al.31 discovered that green innovation can significantly enhance energy efficiency. Research has identified a relationship between green innovation and carbon emissions reduction25,26,32. Some researchers have determined that green innovation is ineffective at decreasing carbon emissions33,34. Cai et al.35 identified a negative link with carbon emissions, resulting in a regional decrease in emissions in the eastern and central areas, while fostering green innovation in the western region. Weina et al.36 asserts that green innovation may possibly improve ecological performance, but has not substantially reduced carbon emissions. The impact of green innovation on CO2 reduction varies significantly among economies37. Chen and Lee38 and Dauda et al.39 found that innovation has reduced CO2 emissions in industrialized nations. However, it has increased them in developing and impoverished nations. The IEA forecast 2021 predicts that green technology, such as renewable energy, will be the front-runner in reducing carbon emissions and maintaining global warming below 1.5 °C40. Zhao et al.’s41 study indicates that technological advancements within low-carbon industries are crucial for reducing emissions in China’s industrial and energy sectors.

The shift to renewable energy is recognized as an essential strategy for achieving environmental sustainability, and has received considerable attention in recent research42,43,44,45,46. The consumption of non-renewable energy significantly contributes to CO2 emissions, with fossil fuels accounting for over 80% of the world’s energy use12. Yang et al.47 asserted that the utilization of nonrenewable energy sources substantially contributes to CO2 emissions. Nevertheless, the global potential for renewable energy is anticipated to increase by 50%, potentially mitigating the detrimental effects of greenhouse gas emissions on the environment48. Numerous studies have shown that renewable energy significantly contributes to reducing carbon emissions, reducing dependence on fossil fuels, and improving environmental outcomes11,49,50,51. For instance, Ref52. a significant correlation was found between the adoption of renewable energy and enhanced environmental quality using the AMG method. Similarly, Ref53. utilized the ARDL framework and concluded that renewable energy significantly reduces emissions. However, varying outcomes have been reported. Bulut54 observed minimal emissions reductions associated with renewable energy. Chen and Lei55 identified a positive relationship between an increase in clean energy utilization and enhanced environmental performance. Inglesi-Lotz and Dogan56 observed that renewable energy substantially reduced emissions. Research has underscored the significance of the sustained deployment of renewable energy to attain carbon neutrality. Countries with increased renewable energy consumption have experienced notable reduction in CO2 emissions, underscoring the importance of continuous investment in green energy innovation57,58. Increasing the share of renewable energy in the national energy mix has a significantly negative correlation with CO₂ emissions4. Despite China’s rapid growth in its renewable energy capacity, coal continues to dominate the power sector, limiting the full potential of renewables4. Renewable energy mitigates carbon intensity, enhances energy composition, and advances industrial structure59.

Environmental taxation has been recognized as a powerful economic instrument for internalizing the external costs of pollution. Environmental taxes are policy instruments designed to direct energy use towards sustainable practices, simultaneously promoting the interplay between ecological conservation and economic growth60. Previous studies have highlighted the correlation between environmental taxes and carbon emissions61,62,63,64. Other studies have similarly indicated the existence of an ideal tax to reduce carbon emissions, proposing that environmental taxes are more effective in decreasing carbon emissions11,65,66. Carbon taxes can significantly reduce emissions, while promoting sustainable economic growth67,68. Consequently, an environmental tax can effectively decrease the carbon emissions generated by different industries and enhance ecological sustainability69,70.

China’s commitment to achieving carbon neutrality by 2060 represents one of the most ambitious climate pledges globally, nevertheless its pathway remains fraught with complexities due to the heavy reliance on fossil fuels, rapid urbanization, and regional economic disparities. Existing studies typically examine financial development, green innovation, renewable energy or environmental taxes in isolation, occasionally integrating the factors within a single provincial‑level, dynamic framework for China. Sub‑national evidence that addresses endogeneity and temporal dynamics is limited; the moderating role of green innovation in the financial development and carbon emissions relationship pathway is rarely tested; and heterogeneity across the emissions distribution and across regions remains underexplored.

This study employs an extended STIRPAT framework to quantify the combined influence of financial development, green innovation, renewable energy, and environmental taxation on CO2 emissions across 30 Chinese provinces from 2007 to 2021. It specifically examines whether green innovation moderates the relationship between financial development and emissions through an interaction effect. The analysis utilizes Fixed Effects Model, the Generalized Method of Moments and Quantile-on-Quantile Regression to map heterogeneous, distributional, and regional effects. By integrating these drivers within a unified dynamic provincial model and introducing the interaction channel between finance and innovation, the research offers a comprehensive assessment of short-run, long-run, and distributional impacts. The findings deliver novel, region-specific policy elasticities to support China’s carbon neutrality pathway, representing one among of the few provincial-level studies to jointly evaluate these factors using this combined methodological approach. This research seeks to resolve the following questions:

Q1: What are the collective impacts of renewable energy, green innovation, environmental taxation, and financial development on carbon emissions in China?

Q2: Does the impact of financial development on emissions vary according to the degree of green innovation in China?

The remainder of this study proceeds as follows: Section "Methodology and data" details the methodology, data sources, and econometric; Section "Empirical results" presents empirical results; Section "Discussion" discusses the findings with particular attention to counterintuitive results, regional variations, and policy implications; Section "Conclusion and implications" concludes with region-specific recommendations; while Section "Limitations and future research" outlines limitations and future research directions.

Methodology and data

Data source and variable description

This study employs a balanced panel dataset comprising 30 Chinese provinces observed over 15 years (2007–2021), yielding 450 total observations. The sample excludes Tibet, Hong Kong, Macau, and Taiwan due to data availability constraints. Provinces are classified into four regions: Eastern (Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, Hainan), Western (Inner Mongolia, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang), Northern (Shanxi, Liaoning, Jilin, Heilongjiang), and Southern (Anhui, Jiangxi, Henan, Hubei, Hunan). Data sources include China Statistical Yearbook, China Energy Statistical Yearbook, China Patent Database, and China Tax Yearbook. The study examined various factors including carbon emissions, green innovation, renewable energy, environmental tax, financial development, urbanization, human capital, economic output. These variables reflect policy-relevant levers of decarbonization in China. Each has established theoretical channels to emissions via scale, composition, and technique effects. See Table 1.

Table 1 Variable definition.
Full size table

Econometric models

Focusing on China’s carbon neutrality goals, this study analyzes the roles of financial development, green innovation, renewable energy integration, environmental tax policies, urban growth, and human capital investments in driving sustainable transitions by extending Dietz and Rosa’s71 STIRPAT model. STIRPAT is a well-known model in environmental economics, developed as a mathematical generalization of the IPAT model (Impact = population * Affluence * Technology) proposed by Ehrlich & Holdren72. The key advantage of STIRPAT over IPAT is that it assumes proportional environmental impacts across the variables73. China’s carbon neutrality goals require capturing nonlinear dynamics beyond traditional drivers. The STIRPAT approach captures the complex nonlinear interactions among variables that influence ecological systems, thus rejecting the assumption of uniform proportional effects74. It enables the evaluation of environmental impacts through socioeconomic and technological factors and the estimation of causal effects75.

The STIRPAT model is mathematically represented as:

$${I}_{it}=\alpha {P}_{it}^{c}{A}_{it}^{d}{T}_{it}^{e}{\varepsilon }_{it}$$
(1)

where I represent the environmental impact; \(\alpha\) is a constant term scaling factor; and c, d, and e denote the elasticity coefficients for population (\({P}_{it})\), affluence \({(A}_{it})\), and technology (\({T}_{it})\), respectively. Term (\({\varepsilon }_{it})\), accounts for the stochastic error term.

While the STIRPAT model has been instrumental in analyzing environmental impacts through the lens of population, affluence, and technology, scholars have raised concerns about the exclusion of essential variables, such as governance quality, socioeconomic inequalities, and regional heterogeneity75,76. This study extends the STIRPAT framework by designating CO₂ emissions as the dependent variable with green technology innovation (GI), renewable energy adoption (RE), environmental taxation (ET), and financial development (FD) serving as core explanatory variables. Urbanization (UR), human capital (HCI), and GDP are incorporated as control parameters11. The extended STIRPAT specification was selected for this research as it overcomes the limitations of the IPAT model by relaxing its proportionality constraint. This facilitates elasticity-based analysis and enables the use of versatile functional forms to capture the nonlinear and heterogeneous nature of emission drivers in China. The framework also provides the capacity to incorporate key moderating factors like policy and institutional variables, thereby offering a more nuanced analysis without sacrificing comparability to prior environmental economics studies. The model can be represented as follows:

$${CO2}_{it}={\alpha }_{0}+{\beta }_{1}{UR}_{it}+{\beta }_{2}{ET}_{it}+{\beta }_{3}{HCI}_{it}+{\beta }_{4}{FD}_{it}+{\beta }_{5}{GDP}_{it}+{\beta }_{6}{GI}_{it}+{\beta }_{7}{RE}_{it}+{\varepsilon }_{it}$$
(2)

Green innovation (GI) acts as a moderating factor in the relationship between financial development (FD) and CO₂ emissions. To evaluate how GI influences this dynamic, Eq. (3) is augmented with an interaction term (\({\beta }_{8}{GI}_{it}*{FD}_{it}\)), capturing the synergistic effect of these variables on emission reduction. The expanded equation is as follows:

$${CO2}_{it}={\alpha }_{0}+{\beta }_{1}{UR}_{it}+{\beta }_{2}{ET}_{it}+{\beta }_{3}{HCI}_{it}+{\beta }_{4}{FD}_{it}+{\beta }_{5}{GDP}_{it}+{\beta }_{6}{GI}_{it}+{\beta }_{7}{RE}_{it}+{\beta }_{8}{GI}_{it}*{FD}_{it}+{\varepsilon }_{it}$$
(3)

Empirical estimation framework

Figure 1 describes the formal econometric modeling process in three main steps. The analysis begins with Step 1, where Pooled Ordinary Least Squares (OLS) regression is applied. Multicollinearity among predictors was assessed to mitigate the risk of biased estimates caused by intercorrelated variables. Following preliminary diagnostics, the analysis proceeds to Step 2, employing two econometric approaches: the Fixed Effects Model, and the Generalized Method of Moments (GMM) to assess long-term relationships and dynamic causal effects. The Fixed Effects Model accounts for unobserved heterogeneity between entities through entity-specific intercepts, while the GMM Model handles the possibility of endogeneity problems and generates more efficient estimates in dynamic panel data contexts. Finally, Step 3 includes a heterogeneity analysis to study variations in effects within various subgroups, making the results valid and generalizable under multiple situations.

Fig. 1
figure 1

Estimation steps.

Full size image

Generalized method of moments (GMM) and fixed effects

Fixed Effects (FE) and Generalized Method of Moments (GMM) models were applied to investigate causal linkages and temporal dependencies among the variables. The FE are used to control for time-invariant provincial heterogeneity using within-entity variation, which is critical in China’s regionally diverse setting. FE provides consistent short-run partial effects under strict exogeneity77. The GMM is employed to address potential endogeneity from dynamic persistence in emissions, reverse causality, and simultaneity among development variables. GMM uses internal lags as instruments, mitigating bias while preserving panel structure. We report Hansen J and related diagnostics to validate instrument validity and limit instrument proliferation78.. Using both FE and GMM provides complementary short-run and bias-corrected dynamic long-run perspectives. This combined approach strengthens the causal interpretation of the results.

Quantile on quantile regression

This study employs the Quantile-on-Quantile Regression (QQR) methodology developed by Sim and Zhou79 to analyze how financial development, green technological innovation, renewable energy adoption, urbanization trends, human capital accumulation, economic growth (GDP), and environmental taxation influence the distributional quantiles of CO₂ emissions. QQR is an advanced technique that evaluates the interplay between the quantiles of a dependent variable and explanatory factors by modeling their conditional relationships across varying quantile levels. The QQR model investigates the relationship between two variables across the different quantiles of their respective distributions. Unlike the traditional quantile of modeling a single conditional quantile, quantile-on-quantile regression estimates the conditional quantiles of one variable as a function of the conditional quantiles of another variable. The QQR model has been widely used to model ecological footprint80,81. QQR is selected to map how the impact of each driver varies across the joint distributions of emissions and the covariate, uncovering asymmetric and regime-dependent effects. This is necessary given China’s heterogeneous provinces and transition stages, where average effects can mask policy-relevant tails. QQR complements mean estimators by revealing where policies are most or least effective.

The QQR model is expressed as follows:

$${C02}_{it}={\delta }^{\tau }+\left({GI}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(4)
$${GI}_{it}={\theta }^{\tau }+\left({CO2}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(5)
$${C02}_{it}={\delta }^{\tau }+\left({UR}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(6)
$${UR}_{it}={\theta }^{\tau }+\left({CO2}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(7)
$${C02}_{it}={\delta }^{\tau }+\left({GDP}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(8)
$${GDP}_{it}={\theta }^{\tau }+\left({CO2}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(9)
$${C02}_{it}={\delta }^{\tau }+\left({FD}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(10)
$${FD}_{it}={\theta }^{\tau }+\left({C02}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(11)
$${C02}_{it}={\delta }^{\tau }+\left({HCI}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(12)
$${HCI}_{it}={\theta }^{\tau }+\left({C02}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(13)
$${C02}_{it}={\delta }^{\tau }+\left({RE}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(14)
$${RE}_{it}={\theta }^{\tau }+\left({C02}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(15)
$${C02}_{it}={\delta }^{\tau }+\left({ET}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(16)
$${ET}_{it}={\theta }^{\tau }+\left({C02}_{t}\right)+{\varepsilon }_{t}^{\tau }$$
(17)

Empirical results

Descriptive statistics

The descriptive statistics in Table 2 summarize the key metrics for a dataset comprising 450 observations across eight variables. Carbon emissions exhibited a mean value of 27,397 with substantial dispersion (standard deviation: 22,729), ranging from 2,321 to 154,180, indicating extensive inter-regional or cross-national divergence in emission intensities. Urbanization levels displayed a mean of 85.671%, indicating a predominantly urbanized demographic profile. Environmental tax revenues show relatively modest mean values (301.962), but high variability (Std: 834.198), reflecting marked heterogeneity in fiscal policy implementation. Human capital and income metrics averaged 23,765 yet spanned a broad spectrum (116–183,980), underscoring pronounced economic stratification. Similarly, the GDP and financial development indicators exhibit considerable economic diversity, as evidenced by their expansive ranges. Green technology (mean: 707) and renewable energy adoption (mean: 533) demonstrated moderate investment levels, although regional adoption patterns were disparate.

Table 2 Descriptive statistics.
Full size table

Figure 2 illustrates significant regional variations in CO2 emissions from 2007 to 2021. The figure displays temporal trends across 30 provinces, with substantial heterogeneity in both emission levels and trajectories. Shandong province, reaching a peak of 154,180 kt in 2021, shows a 58% rise from 2007 levels, driven by coal dependent industries and rapid industrial expansion. This province represents the highest emitter throughout the sample period, reflecting its status as China’s major industrial hub. In contrast, Qinghai (minimum: 2,321 kt) has stable patterns attributable to its reliance on hydropower and low industrial intensity, remaining among the lowest emitters throughout the period. Many provinces show rising emissions through 2012 and 2013, coinciding with China’s post-2008 stimulus period that accelerated infrastructure development and industrial expansion. This upward trend then flattens or declines after 2016, corresponding to the implementation of China’s 13th Five-Year Plan (2016–2020), which introduced stricter environmental regulations and industrial restructuring mandates. Six provinces, including Hebei and Inner Mongolia, exhibited marked instability from 2015 to 2017, with temporary emission spikes occurring as coal facilities increased output before closure. Declines followed subsequent to renewable energy subsidies and policy enforcement. These inconsistencies highlight the impact of national policies on localized emission responses, necessitating provincial-level monitoring in our models.

Fig. 2
figure 2

Source: authors’ computation.

CO2 emissions (kt) in 30 Chinese Province from 2007 to 2021.

Full size image

Multicollinearity test

Multicollinearity test was performed to confirm regression robustness. Variance Inflation Factors (VIF) revealed moderate multicollinearity for GDP (VIF = 6.23) and mild multicollinearity for other variables (VIF < 5). The mean VIF (3.07) confirmed no significant collinearity issues, as it fell below the conservative threshold of 582. All subsequent models are therefore statistically reliable.

Estimation results

Table 3 summarizes the regression outputs from the Fixed Effects (FE) and Generalized Method of Moments (GMM) frameworks. The FE model showed a significant positive relationship between UR and CO2 emissions (96.65** to 101.0**), suggesting that UR increases CO2, possibly due to higher energy demand in urban areas. Conversely, the GMM estimates are negative and significant (− 29.78** to − 30.05**), implying that urbanization may contribute to emission reductions when accounting for endogeneity and dynamic effects. These findings support83 the possibility that emissions could be decreased.

Table 3 Fixed effects and GMM results.
Full size table

The environmental tax in FE is insignificant (coefficient ~ 1.0), suggesting that the short-term effects are unclear. The GMM estimates show a significant negative relationship (− 0.376*** to − 0.361***), suggesting that sustained environmental tax policies contribute to long-term carbon emission reductions. This finding highlights its effectiveness in reducing emissions when accounting for policy lags and endogeneity. Taxes may incentivize cleaner technologies over time.

GDP is negative but insignificant in FE (− 0.0101 to − 0.0467), indicating that economic growth does not exhibit a clear directional relationship with CO2 emissions in the short term. Economic growth may increase emissions through industrial activities but also enable investments in green technologies, decreasing CO2 emissions84. As the economy grows rapidly, it tends to produce more CO2, stressing the environmental costs associated with economic growth. This implies that the increased industrial activity and energy consumption linked to economic expansion can increase CO2 emissions levels85. These findings are similar86,87. GMM is negative and significant (− 0.00981** to − 0.00847*), indicating that economic growth is weakly associated with reduced emissions over the long term, aligning with the Environmental Kuznets Curve (EKC) hypothesis, where economies eventually decouple growth from emissions through structural shifts88,89,90.

FE shows a positive and significant relationship between HCI and CO2, with coefficients of 0.250* (10% significance) and 0.325** (5% significance). This finding indicates that higher human capital correlates with increased CO2 emissions in the short term. Human capital boosts productivity and consumption, temporarily increasing emissions. The GMM was insignificant (0.0261 and 0.0241, respectively). This indicates long-term offsetting effects. While skilled populations may drive industrial growth and increase emissions, they may also foster green innovation or demand for sustainability over time. The results support the conclusions that human capital initially reduces CO2 emission intensity, subsequently raises it in the short term, and eventually lowers it over the long run91,92.

The FE models show a significant positive relationship between financial development (FD) and carbon dioxide (CO₂) emissions (0.257*** to 0.284***), suggesting that financial growth initially correlates with higher emissions. The findings support the argument that, as the financial system grows, it often relies more on industries that produce carbon emissions. This growth can lead to increased consumption of fossil fuels and higher emissions. The findings align with Refs28,93,94,95.. On the other hand, GMM is negative and significant (− 0.0584***), indicating that mature financial systems can significantly contribute to combating environmental degradation by funding cleaner technologies and sustainable practices. The findings are consistent with96,97 previous research, which indicates that financial development contributes to lower carbon emissions.

FE is negative in one model (− 1.453*) but insignificant in another, hinting at potential short-term reductions. These findings are similar to32,98 showing that GI reduces CO2 emissions. The GMM models are positive and significant (0.521*** to 0.442***). While innovations may be green, the overall industrial context could still lead to increased emissions, possibly due to production and consumption activities. The results align with Suki et al.99 endogenous technological innovation raises CO2 emissions due to the economic scale effect, while simultaneously decreasing CO2 emissions by enhancing efficiency.

The positive relationship between renewable energy (RE) and CO₂ emissions (7.631*** in FE; 0.563** in GMM) appears counterintuitive but reflects significant contextual nuances. Quantile regression (Fig. 4xi) shows RE reduces emissions only at higher adoption quantiles. In provinces with low RE adoption, renewables operate as a supplement rather than a substitute for fossil fuels, failing to displace coal dominated baseload. Moreover, the carbon footprint of renewable manufacturing may offset short-term gains. Similarly, the positive coefficient of green innovation (GI) in GMM models (0.521***) stems from the scale effect. Innovation boosts production efficiency, lowering costs and increasing consumption and emissions before net reductions materialize. These findings emphasize that RE and GI require scale, temporal persistence, and fossil fuel displacement to achieve carbon reduction.

Heterogeneity analysis

Regional heterogeneity is fundamental for understanding China’s decarbonization pathway and constitutes a pivotal finding of this analysis. The effects of financial development, green innovation, renewable energy, and environmental taxation vary significantly in direction and scale across regions, underscoring the profound economic, geographic, and institutional diversity. These disparities demonstrate that uniform national policies are inadequate. The variation in how urbanization, financial development, and green innovation influence CO₂ emissions is essential for formulating effective, region-specific strategies.

Tables 4 and 5 present the results for Models 1 and 2 for regional fixed effects estimations, which reveal substantial regional variations in emission drivers. In the eastern region, urbanization intensifies emissions, with a coefficient of 643.6, a result of energy-intensive industrialization and infrastructure development. Conversely, the northern region shows emissions driven by heating demand and transportation, with a coefficient of 275.7. The western region exhibits no significant urbanization impact, likely due to low population density and greater reliance on renewable energy sources.

Table 4 Model 1 fixed effects results for regions.
Full size table
Table 5 Model 2 fixed effects model estimation results for regions.
Full size table

ET is counterproductive in the east, with a coefficient of 128.5, where industries treat taxes as an operating cost without reducing output. However, it proves effective in the south, with a coefficient of 29.72. FD intensifies emissions in the east, with a coefficient of 2.400, but reduces them in the north, with a coefficient of -0.396, where mature financial systems fund green projects. The interaction between FD and GI is significant only in the east, with a coefficient of − 0.00117, and the south, with a coefficient of 0.000263. This indicates that FD enhances the emission-reduction effects of green innovation in advanced eastern economies but may fund greenwashing in southern industrial hubs.

These sharp discrepancies require distinct policy actions for each region. Eastern provinces require output based environmental regulations that mandate absolute emission reductions, moving beyond cost-based taxes. Policy should also direct FD toward certified green technologies. Southern regions need strict oversight and verification protocols for green innovation projects to prevent greenwashing. Northern provinces should continue leveraging financial systems for clean heating transitions and transportation electrification. Western China benefits from policies that maintain its renewable energy trajectory and prevent future carbon-intensive lock-in.

Correlation matrix

Figure 3 shows a moderate positive relationship between CO2 and UR, RE, FD, and GI. The correlation between ET and CO2 emissions is relatively low. There is also a positive correlation between RE, GDP, HCI, FD, GI, and UR, while there is a low correlation between ET and UR. There is a strong correlation between GDP, GI, HCI, and RE, whereas FD has a moderate connection. ET has a negative correlation with RE. The independent variables of GI, HCI, and FD exhibited strong correlations with GDP, whereas the correlation between ET and GDP was negative. The correlation between ET and HCI was negative, whereas that between FD, GI, and ET was weak. GI, FD, and HCI were positively correlated. There is a positive association between FD and GI.

Fig. 3
figure 3

Source: Author’s computation.

Correlation matrix.

Full size image

Quantile on quantile regression analysis

The empirical findings of the QQR analysis are presented in Fig. 4, which displays the plots of the slope coefficient. Figure 4i shows the relationship between carbon emissions and the GDP. These trends reveal that more significant carbon emissions correlate with economic growth, especially in lower GDP quantiles. This implies that carbon-intensive sectors drive economic growth in relatively poor economies. At advanced GDP levels, the effect weakens, reflecting that more prosperous economies might have already embraced cleaner technologies or reduced carbon dependency.

Fig. 4
figure 4figure 4figure 4

(i) CO2 on GDP, (ii) GDP on CO2, (iii) CO2 emissions on ET, (iv) ET on CO2 emissions, (v) CO2 on UR, (vi) UR on CO2, (vii) CO2 in GI, (viii) GI in CO2, (ix) FD on CO2, (x) CO2 on FD, (xi) RE on CO2, (xii) CO2 on RE, (xiii) CO2 on HCI, (xiv) HCI on CO2.

Full size image

Figure 4ii indicates that GDP initially increases CO2, particularly in the lower to mid quantiles (0.2–0.6). However, the effect decreases at higher quantiles, possibly because of the improved GI and more stringent environmental regulations in advanced economies. Figure 4iii depicts the influence of CO2 on the ET policy. Although emissions are more significant at lower quantiles, this does not necessarily result in immediate, stringent taxation. However, ET is steep for higher quantiles, where emissions are extremely high, indicating a policy measure to curtail emissions.

Figure 4iv shows the impact of ET on CO2. These shapes suggest that ET effectively reduced emissions in the lower and higher quantiles, confirming that policies can restrict pollution. However, this does so to a lesser degree in mid-level quantiles, which may indicate inefficiency in tax implementation or industrial adjustment. Figure 4v shows the relationship between the UR and CO2. Lower UR levels were correlated with high emissions, likely because of underdeveloped infrastructure and inefficient energy consumption patterns. Conversely, higher UR showed an inverse correlation with emissions, potentially owing to improved energy efficiency, urban planning, and public transportation networks.

Figure 4vi suggests that UR contributes to emission reduction across both low and high quantiles, supporting the notion that urbanized areas can mitigate their carbon footprints. However, the effect is ambiguous in the mid-quantiles, possibly because of complex development trends and mixed urban policies. Figure 4vii shows that CO2 are still high with low levels of green innovation. However, with increased levels of green innovation, emissions start to decline, corroborating the assertion that investment in clean technologies is paramount in ensuring that emissions fall over the long run.

Figure 4viii value reinforces the role of GI in reducing CO2, and more so in higher quantiles, where more established solutions to innovative issues are embraced. The reduced impact at lower quantiles implies that it could take a while for early investment in green innovation to reduce emissions. Figure 4ix indicates that FD correlates with increased emissions in the lower quantiles, potentially due to investments in energy-intensive sectors. However, at higher quantiles, the relationship goes in the opposite direction, which implies that advanced financial systems stimulate cleaner energy investment.

Figure 4x shows that increased CO2 can inhibit FD at higher quantiles, because investors tend to invest in sustainable sectors. Nevertheless, this relationship was weak at lower quantiles, suggesting that environmental issues might not be a concern in the early stages of FD. The curve in Fig. 4xi implies that adopting RE lowers emissions substantially at higher quantiles. The lower impact at lower quantiles means that adoption in the early stages may not be sufficient to achieve significant reductions. Figure 4xii shows that high CO2 seem to hasten the shift towards RE at higher quantiles, possibly because of policy measures and growing awareness. However, the impact is unclear in the lower quantiles, indicating that RE adoption can still be hindered in the initial phases.

Figure 4xiii shows that economies with lower HCI have more significant emissions, perhaps because they depend on conventional energy sources. With increasing HCI (higher quantiles), emissions were reduced, reflecting the significance of education and skills in sustainability. The estimate in Fig. 4xiv is consistent with the idea that HCI development results in lower emissions, especially in the high quantiles. The minor effect in the mid-quantiles implies that policy and infrastructure are also determinants of emission reduction.

Discussion

The findings reveal the dynamics influencing China’s carbon emissions, with profound implications for policymakers. Urbanization initially intensifies emissions driven by China’s unprecedented construction surge and increasing private vehicle ownership but ultimately mitigates them as cities develop. Early urbanization relies on carbon-intensive materials and fragmented transit systems, while developed cities leverage economies of scale for metro networks, energy-efficient buildings, and cohesive land-use planning. The findings align with13,14,15 projections that China’s carbon peak requires structural urbanization adjustments,shifting from sprawl-centric growth to compact, transit-oriented development models.

The impact of green innovation on CO₂ emissions exhibits an intricate temporal pattern. Our empirical results show statistically mixed short-run effects in the Fixed Effects models, with one specification indicating a reduction (-1.453) and another showing an insignificant relationship. Conversely, the long-run Generalized Method of Moments estimates reveal a positive association, with coefficients ranging from 0.442 to 0.521. This pattern is consistent with the existing literature, which underscores context-dependent outcomes. While Ref13,14,15. emphasize innovation’s vital role in the carbon neutrality pathway, and Cai et al.35 document regional emission reductions in strong innovation clusters, our findings suggest a potential rebound effect in early phases, where efficiency gains may not immediately translate into absolute emission reductions at the aggregate level. This difference is critical because it demonstrates that patent counts alone are an insufficient metric for guaranteed reduction. Therefore, an essential policy lever involves conditioning innovation support on verified emission reductions, achieved through firm-level monitoring, reporting, and verification systems to ensure tangible climate benefits.

The estimated impact of environmental taxation on CO₂ emissions reveals a significant temporal dimension. The short-run Fixed Effects coefficients are statistically insignificant, indicating a limited immediate effect. In contrast, the long-run Generalized Method of Moments estimates show a significant negative relationship. This pattern suggests the operation of a lagged enforcement channel, where the full effect of the policy materializes over time. Firms initially perceive environmental taxes as an additional operational cost to be absorbed. In the long run, however, they respond by undertaking structural and technological changes to mitigate this financial burden. This finding is consistent with the established literature. Studies by Xie et al.67 and Zhang et al.68 demonstrate the emission reduction efficacy of well-designed tax policies, while Li et al.70 explicitly links regulatory stringency to the adoption of green technologies. Therefore, the critical policy lever extends beyond merely setting a tax rate. Its effectiveness hinges on establishing credible enforcement and regular audits to ensure compliance. Furthermore, earmarking the generated tax revenues for green capital formation and technological subsidies can create a virtuous cycle, accelerating the transition to a low-carbon economy.

The relationship between renewable energy adoption and CO₂ emissions reveals a critical complexity in the current context. Empirical estimates show a positive relationship in both the short-run and the long-run. This finding is further refined by the Quantile-on-Quantile Regression, which indicates emission reductions only materialize at the highest quantiles of renewable energy adoption. In coal-dependent provinces, analysis suggests that a one percent increase in renewable capacity is associated with an approximate 0.2% rise in emissions. This counterintuitive result aligns with studies such as Ahmed et al.57, Chen (2023), and Lee (2023), which generally report negative emission effects from renewable expansion. The collective evidence suggests that at its current scale and within the existing infrastructure, renewable energy primarily supplements, rather than shifts, fossil fuel consumption. This distinction is critical because it highlights the persistent reliance on coal in regional grids and industrial processes. Consequently, effective policy must move beyond mere capacity addition to integrate the phased retirement of coal power plants, strategic investment in storage technologies, and the modernization of transmission grids to ensure renewables replace fossil fuels rather than complement them.

Financial development (FD) shows a temporal duality: it initially contributes to emissions by financing fossil fuel projects, but thereafter facilitates decarbonization through green bonds and venture finance. The findings align with Ref25,26., green finance facilitates emission reductions. This pattern suggests the dominance of scale effects in the short term, where financial expansion fuels economic activity and emissions, giving way to technique effects in the long run, where mature financial systems facilitate decarbonization. Consequently, the critical policy implication involves directing credit allocation toward credibly verified green projects to accelerate this transition and harness the long-run decarbonization potential of financial development.

Conclusion and implications

This study analyzes the factors influencing carbon emissions, focusing on urbanization, environmental taxes, GDP, human capital, financial development, green innovation, and renewable energy, using panel data from 30 Chinese provinces from 2007 to 2021. The findings show the varying impacts of different factors on CO2 emissions. Urbanization initially increases emissions, but may lead to long-term reductions when considering endogeneity, suggesting that policymakers should prioritize sustainable urban planning and development strategies that leverage the enduring advantages of urbanization while mitigating short-term environmental costs. These findings underscore the efficiency of environmental taxes in reducing CO2 emissions, specifically in the long term. Emphasizing the significance of consistent and effectively structured environmental tax policies to achieve carbon neutrality goals. Policymakers should consider implementing and maintaining such policies while addressing their potential short-term economic impacts. The relationship between GDP and CO2 emissions aligns with the Environmental Kuznets Curve hypothesis, suggesting a possible decoupling of economic growth from emissions in the long term. This implies that continued economic development, investments in cleaner technologies, and structural shifts towards less carbon-intensive industries could reduce emissions without sacrificing growth. The study reveals that human capital has short-term positive effects on emissions, but insignificant long-term effects. Targeted education and training programs emphasize environmental awareness and sustainable practices to leverage human capital for long-term emissions reduction.

Although financial development may initially increase emissions, established financial systems can facilitate a decrease in emissions. This implies that policymakers should promote the development of green financial products and incentivize investments in sustainable projects to accelerate the transition to a low-carbon economy. The mixed effects of green innovation on emissions highlight the complexity of technological progress in the environmental context. This finding suggests the need for more focused and effective green innovation policies that foster the development of clean technologies and guarantee their extensive adoption and incorporation into current industrial practices. Contrary to expectations, this study found a positive association between renewable energy and CO2 emissions. This counterintuitive result warrants further investigation, and may imply that the current scale of renewable energy adoption is insufficient to offset emissions from other sources. Policymakers should consider accelerating the transition to RE while addressing the emissions from different sectors. The insignificant interaction term suggests that financial development’s effect on emissions is not systematically moderated by green innovation levels in the short or long term. Provinces with higher GI do not exhibit a statistically distinct relationship between FD and emissions compared to provinces with lower innovation. Policymakers should still pursue integrated strategies to align financial systems with green innovation, as both independently influence emissions.

The heterogeneity analysis revealed significant regional variations in the effects of various factors on CO2 emissions. Underscore the importance of tailored region-specific policies that account for local economic, social, and environmental conditions. Quantile regression analysis revealed complex nonlinear relationships between various factors and CO2 emissions across different emission levels. For instance, the impacts of urbanization, financial development, and human capital on emissions appear to change over time or at varying levels of development. This suggests that policies aimed at reducing emissions may need to be adaptive and context-specific. Additionally, the study underscores the importance of considering both short- and long-term effects, as some factors that initially increase emissions may lead to reductions over time as technologies and practices evolve. These findings emphasize the need for comprehensive long-term strategies that simultaneously address multiple factors to effectively reduce CO2 emissions.

Drawing on our findings, we recommend tailored strategies for China’s major regions. In Eastern China, where rapid industrialization and urbanization drive high emissions, policymakers should implement stringent emissions caps coupled with green finance incentives. Preferential loans for renewable energy projects and subsidies for commercializing green patents would leverage the significant FD and GI synergy identified in our models. This approach counterbalances scale effects from urban expansion while accelerating technology diffusion in high-emission industrial clusters. For Northern China, the dominant influence of heating demand and transportation on emissions necessitates transitioning district heating systems from coal to solar sources and expanding electrified public transport networks. Redirecting financial development (FD) toward modernizing heating infrastructure evidenced by FD’s negative emission elasticity would capitalize on the region’s developed financial system. Western China requires scaling utility-scale renewable energy infrastructure to move beyond token deployments, a critical response to our finding that RE only reduces emissions at high adoption quantiles. Environmental tax collection mechanisms should be reformed to penalize non-compliance transparently, addressing observed inefficiencies. Skill-building programs aligning human capital with green sectors are essential to activate the region’s underutilized emission-reduction potential. In Southern China, environmental tax reforms must enforce strict penalties for non-compliance while rewarding verified emission reductions, countering the counterproductive tax-emission linkage. Rigorous auditing of FD and GI projects is imperative to prevent greenwashing, particularly given the positive interaction term suggesting financial resources may inadvertently fund superficial sustainability initiatives.

Limitations and future research

This study is subject to several limitations. First, the data primarily originate from provincial statistical yearbooks, and potential inconsistencies in reporting standards may introduce measurement error. Second, the exclusion of Tibet, Hong Kong, Macau, and Taiwan limits the geographic coverage and generalizability of findings, though this exclusion was necessitated by data availability constraints. Third, the 2007 to 2021 window omits recent policy shifts after the COVID-19 pandemic, which may have altered the relationships identified. The pandemic’s impact on economic activity, supply chains, and policy priorities represents a structural break that future research should address. Fourth, the renewable energy variable may reflect installed capacity rather than actual fossil fuel displacement in electricity dispatch. The green innovation is based on patent counts and does not necessarily indicate technology adoption. Similarly, the environmental tax variable captures statutory rates and revenues but not enforcement intensity. Results apply specifically to Chinese provinces during the 2007–2021 period and may not generalize to other countries or time periods. The regional groups we employ (Eastern, Western, Northern, Southern) are broad and may mask important within region heterogeneity. The findings are particularly relevant for rapidly industrializing economies with heavy coal dependence, but may differ in contexts with different energy structures, financial systems, and governance arrangements. The fixed effects control for time-invariant provincial characteristics and GMM addresses endogeneity, residual bias may persist from weak instruments or correlated provincial shocks. The Quantile-on-Quantile analysis describes distributional relations but does not establish causal structural breaks.

Future studies could advance this research agenda in several directions. Combining official statistics with satellite-based emissions data and power plant generation records would enhance measurement accuracy. Research should also link green finance indicators to firm-level reduction outcomes to clarify the mechanisms behind the financial development and green innovation relationship. Integrating infrastructure factors such as coal retirement schedules, storage capacity, and transmission grid constraints into models would help quantify the conditions under which renewables effectively replace fossil fuels. Employing quasi-experimental designs to evaluate specific environmental tax reforms would strengthen causal inference. Further investigation into non-linear thresholds and multi-way interactions between financial development, green innovation, renewable energy, and environmental taxation could identify effective, region-specific policy packages. Comparative studies examining similar relationships in other large developing economies (India, Brazil, Indonesia) would help distinguish China-specific findings from broader patterns in rapidly industrializing nations. Finally, extending the panel analysis beyond 2021 is crucial for capturing the effects of recent economic recovery and intensified climate policies.