How Does Carbon Emissions Trading Impact Energy Transition? A Perspective Based on Local Government Behavior

Abstract

Assessing the environmental and economic impacts of the carbon emissions trading scheme (ETS) is both timely and essential. This study investigates the effects of the ETS on energy transition by analyzing panel data from 30 provinces and municipalities across mainland China. The findings highlight three key points. First, the ETS significantly promotes energy transition. Robustness tests confirm the validity of this conclusion. Compared with non-pilot provinces, pilot provinces achieve a 4.83% increase in energy transition levels. Second, the energy transition effect of the ETS is mainly achieved by changing the incentive and constraint behavior of local governments. Third, the ETS exerts a more pronounced impact on energy transition in regions with higher levels of marketization and stronger innovation capabilities. Furthermore, the effects of the ETS vary across different quantiles of energy transition levels. This study provides a novel perspective on achieving the synergistic development of economic growth and environmental sustainability.

1. Introduction

The climate crisis represents one of the most pressing global challenges of our time, prompting the international community to cooperate actively in addressing climate change. In 2016, the Paris Agreement came into effect, establishing the goal of temperature rise control in this century. Against this backdrop, countries worldwide have successively announced carbon neutrality targets. As a leading global actor, China has emphasized tackling climate change and committed to reaching carbon neutrality by 2060.

Since the 1920s, the energy industry has been the leading contributor to human-generated carbon emissions [1]. According to the International Energy Agency, energy generation and heating are the largest contributors to global carbon emissions. Consequently, achieving substantial reductions in carbon emissions necessitates a fundamental transformation of global energy systems. China is the largest developing country in the world with huge energy consumption. Between 1980 and 2020, China’s energy consumption increased from 586 million to 4.557 billion tons of standard coal equivalent, reflecting a consistent upward trend. However, China’s energy structure remains heavily reliant on coal, which constitutes nearly 70% of its total energy consumption—2.4 times the global average [2]. This dependence on coal has led to persistently high carbon emissions, intensifying international pressure on China to address climate change and posing significant risks to energy security [3]. As a result, accelerating the energy transition and achieving coordinated economic and environmental governance have become critical priorities for China.

Environmental regulation is a critical instrument for encouraging entities to undertake environmental protection measures [4,5,6]. The ETS, as a market-based environmental regulation, aims to control carbon emissions by pricing carbon and establishing corresponding trading markets. The ETS is widely recognized as a key mechanism for reducing carbon dioxide emissions and mitigating global warming [7,8]. According to the World Bank, 31 countries or regions have implemented or plan to implement ETSs, while over 160 countries or regions have set carbon neutrality targets [9,10]. In this context, China’s National Development and Reform Commission approved carbon trading pilot programs in seven provinces and municipalities in 2011. Beginning in 2013, these carbon markets were gradually launched, and China’s ETS is progressing steadily toward standardization and internationalization.

Energy transition is essential for advancing environmental governance and achieving high-quality economic development. However, can the ETS effectively promote energy transition in China? Does its impact vary across regions? What are the underlying mechanisms of influence? This study explores these questions by integrating theoretical frameworks with data analysis.

Key findings from this research include the following: First, it develops a multi-dimensional evaluation index system to measure the level of energy transition, addressing the limitations of previous studies that often relied on single indicators. This approach also mitigates potential endogeneity issues in subsequent causal analyses. Second, this study evaluates the impact of China’s ETS on energy transition, providing an objective and accurate quantitative analysis to support policy evaluation. Third, it investigates the behavioral motivations of local governments in implementing the ETS and analyzes the mechanisms through which their incentive and constraint behaviors influence energy transition, thereby enriching the theoretical research on environmental regulation policies. Finally, it examines the heterogeneous effects of the ETS on energy transition from the perspectives of regional innovation capacity and marketization levels, offering insights to enhance the ETS and draw lessons from carbon markets.

This research is organized into seven distinct sections. To begin, Section 2 delves into a comprehensive review of existing literature. Section 3 then lays out the theoretical foundation underpinning this study. Moving forward, Section 4 provides a detailed account of the data sources and methodological approach employed. Section 5 presents the empirical results, offering insights into the findings. Section 6 takes a closer look at the underlying mechanisms at play. This paper wraps up in Section 7, drawing conclusions and highlighting key policy recommendations.

2. Literature Review

2.1. Research on the Carbon Emissions Trading Scheme

As a pivotal instrument within environmental regulatory frameworks targeting emission mitigation, carbon emissions trading schemes (ETSs) have garnered considerable scholarly interest. Recent studies in Europe and North America have further advanced our understanding of ETS effectiveness from both environmental and economic perspectives. In the context of the European Union Emissions Trading System (EU ETS), Zaklan et al. (2021) found that the tightening of Phase IV cap settings led to a significant decline in carbon emissions from the power and industrial sectors [11]. Similarly, Brunink et al. (2020) emphasized the role of market stability reserves in reducing allowance oversupply and strengthening carbon pricing mechanisms [12]. In the United States, Lessmann et al. (2024) evaluated California’s Cap-and-Trade Program and demonstrated its effectiveness in reducing emissions without adversely affecting industrial output [13]. These findings reinforce the importance of market design, allowance scarcity, and policy credibility in determining ETS performance. Comparative insights from mature ETS markets offer valuable lessons for emerging systems like China’s, particularly in areas such as auction mechanisms, cross-sectoral coordination, and carbon leakage prevention.

Since the establishment of China’s carbon emissions trading pilots, a number of studies on the impact of China’s carbon emissions trading system have begun to emerge, but mainly focus on energy conservation and emission reduction. Hu et al. (2020) and Chen et al. (2021) demonstrated that China’s ETS plays a critical role in reducing carbon dioxide emissions [14,15]. Similarly, Qi et al. (2020) showed that the ETS effectively reduces carbon emissions without impeding economic growth [16]. In addition to confirming the effectiveness of the ETS, Xuan et al. (2020) also proposed factors that affect the process of carbon emission reduction [17]. Nevertheless, not all findings are unequivocally positive; some analyses suggest unintended consequences, such as energy rebound effects and the phenomenon of carbon leakage [18,19].

2.2. Literature on the Concept and Dimensions of Energy Transition

In 1982, the concept of energy transition was introduced in Germany [20]. Since then, academic discussions on energy transition have steadily evolved. Our understanding of this concept has deepened over time, progressing from surface-level interpretations to more comprehensive insights [21]. Initially, scholars regarded energy transition as a shift in energy structures, focusing on the transition in energy production and consumption from non-renewable sources to renewable energy [22]. For instance, Hefner III (2009) revealed the evolution path of energy transition by classifying the three forms of energy: solid, liquid, and gas [23]. Fan and Yi (2021) emphasized that the core of energy transition lies in the fundamental change of energy supply-side structure [24]. According to Ma et al. (2018), the concept of energy transition encompasses shifts in energy composition, with a core element being the rising adoption of renewables [25]. Liu et al. (2022) and Qi & Li (2018) also agree with this view [3,26]. As research advanced, the concept expanded from a focus on energy structure transition to encompass the broader transition of energy systems. Given the realities of China’s energy landscape, some scholars defined energy transition as a comprehensive and systemic transition of all components within the existing energy system [27,28,29,30,31,32].

2.3. Research on the Impact of ETS on Energy Utilization

Numerous studies have discussed the impact of ETS on energy utilization. For instance, Liu and Zhang (2021), using Chinese provincial data from 2004 to 2019, demonstrated that the ETS stimulates the development of the non-fossil energy sector [33]. Tan et al. (2022) argued that the ETS lowers energy demand and refines the energy mix by separating economic expansion from reliance on fossil fuels and resultant carbon output [9]. Geng and Fan (2021) showed that the ETS significantly lowered energy intensity in Hubei Province between 2015 and 2017 [8]. Hong et al. (2022) utilized a DID approach to demonstrate that ETS boosts energy efficiency by fostering technological advancement [34]. Meanwhile, Borghesi et al. (2015) and Tan et al. (2022) noted that the effect of ETS on energy efficiency are not uniform, differing significantly among various companies and geographical areas [35,36]. However, Chen et al. (2021) found that the energy-saving effects of the ETS have fallen short of expectations [18].

The factors influencing energy transition have been a central focus of academic research. Economic development, technological advancement, human capital, and foreign investment are widely recognized as key drivers of energy transition [37,38,39] (Pagliuca et al., 2022; Shahbaz et al., 2022; Qi et al., 2022). However, limited research has explored the effects of ETS on energy transition or analyzed the particular mechanisms by which China’s ETS affects energy transition.

3. Theoretical Framework and Hypotheses

3.1. Construction of the Energy Transition Evaluation Index System

Energy transition is a systematic project involving the economy, energy, environment, technology, and other fields. It is difficult to fully reflect its effectiveness and challenges by a single-dimension evaluation. Through multi-dimensional comprehensive evaluation, we can more comprehensively grasp the progress and problems of energy transition and its comprehensive impact on social economy, and provide a scientific basis for policy formulation and implementation [40,41]. Among the most comprehensive tools for evaluating energy transition is the World Economic Forum’s Energy Transition Index (ETI). This index not only evaluates national energy system performance, but also considers energy transition readiness through macroeconomic, institutional, social, and geopolitical dimensions [41]. As of 2024, the ETI has assessed the energy transition performance of 120 countries, offering a clear benchmark for global progress [42]. While the dimensions of energy transition are broadly similar across countries, unique national contexts must be carefully considered in such assessments. Therefore, advancing research on energy transition at the sub-national level should be a priority.

Drawing on the ETI and guided by the principles of systematicity, hierarchy, scientific rigor, and data accessibility, this study develops a provincial-level energy transition evaluation framework. The framework encompasses five dimensions: energy, environment, economy, technology, and human capital (

Table 1. Energy transition evaluation index system.

Evaluation Dimension	Element Layer	Indicator Layer	Index Attribute
energy	energy structure	coal proportion	-
	energy consumption	per capita energy consumption	-
	energy intensity	energy consumption per unit GDP	-
environment	air pollution	average PM2.5 concentration	-
	carbon intensity	carbon dioxide emissions per unit of GDP	-
	carbon emissions	carbon emissions per capita	-
economy	economic development and growth	per capita GDP	+
		GDP growth rate	+
	capital and investment	energy expenditure per capita	+
	economic structure	share of secondary industry	-
		share of the tertiary sector	+
technology	innovation capacity	number of green patent applications per capita	+
	technology expenditure	R&D investment intensity	+
	infrastructure	length of natural gas pipeline	+
human capital	educational quality	per capita expenditure on education	+
		number of students in higher education	+
	training capacity	proportion of employees in the education industry	-

). The weights of each sub-indicator are determined endogenously through the RAGA-PPM optimization process, which identifies the projection direction that maximizes data structure information. This data-driven weighting approach avoids potential biases associated with expert judgment or equal weighting, and enhances the objectivity and robustness of the composite ETL index.

3.2. Theoretical Analysis of the Impact of ETS on Energy Transition

As a market-based environmental governance tool, ETS can be viewed as a strategy for local governments to balance economic growth with environmental protection. Historically, China’s performance evaluation system for officials prioritized GDP, which limited incentives for local officials to prioritize environmental protection. However, with increased emphasis on ecological civilization and sustainable economic development, the evaluation criteria for officials have shifted from a singular focus on GDP to a more comprehensive framework that includes green GDP and environmental quality [43]. The primary objectives of ETS are aligned with the broader framework of environmental economics, particularly the Pigovian approach, which internalizes negative externalities through market-based instruments such as carbon pricing. Moreover, from the perspective of policy diffusion theory, the ETS can be interpreted as a policy innovation that spreads through regional experimentation, institutional mimicry, and adaptive learning across provinces, especially in the context of China’s decentralized environmental governance structure. By advancing green technological innovation and developing the renewable energy industry, local governments can reform regional energy consumption and industrial structures. These efforts address resource and environmental challenges, reduce pollution, improve environmental quality, and strengthen the ‘ecological civilization’ brand. Additionally, the competitive advantages driven by green technological innovation and the renewable energy industry can stimulate employment, boost urban economic growth, and enhance the competitiveness of regional green development [44]. In summary, local governments are strongly incentivized to effectively implement the ETS. Hence, the subsequent hypothesis is:

Hypothesis 1:

The effective implementation of ETS improves regional energy transition.

Enterprises often fail to account for negative externalities in their decision-making processes. Therefore, government intervention in the energy transition sector is essential to address market failures, primarily through incentive and constraint behaviors [45]. This categorization draws upon instrument typologies in public policy theory, which distinguish between tools that encourage behavior through positive reinforcement (incentives) and those that deter through costs or penalties (constraints) [46]. Under this framework, incentive behaviors refer to subsidies, fiscal transfers, and investment support, while constraint behaviors include regulations, caps, and penalties.

For the purposes of this study, incentive behaviors are divided into two aspects. First, investment guidance: local governments play a critical role in policy-driven incentives, directing social investment toward renewable energy production and energy transition enterprises. The enhancement of financing platforms further amplifies this effect [47]. Additionally, market mechanisms attract diverse investment entities and demand-side participants to the renewable energy sector, enriching resources for green technological innovation and fostering firms’ intrinsic motivation for innovation [48]. Second, fiscal incentives: the development of renewable energy generates significant positive externalities, such as reducing pollutant emissions, improving ecological environmental quality, promoting technological innovation, and creating spillover effects. However, these benefits are often inadequately incentivized by market mechanisms. Moreover, renewable energy infrastructure and technological research and development exhibit characteristics of public goods, leading to market underinvestment. To address this, governments intervene in renewable energy development through fiscal subsidies, tax incentives, and policy support. These measures internalize the environmental costs of traditional energy, mitigate underinvestment in renewable energy markets, and drive technological progress. Consequently, such interventions optimize resource allocation, enhance environmental quality, and strengthen energy security and sustainable development capacity at the national level [49]. From this, the following hypothesis is formulated:

Hypothesis 2:

ETS increases fiscal support by influencing the incentive behaviors of local governments, thereby improving regional energy transition.

In terms of constraint behaviors, ETS effectively promotes energy transition through dual constraints on energy intensity and carbon intensity. On the one hand, ETS imposes a cap on total carbon emissions and progressively tightens the allocation of allowances, compelling firms to optimize energy efficiency and reduce carbon emission costs, thereby driving a decline in energy intensity [50]. Facing the challenge of rising carbon emission costs, industrial entities should focus on implementing energy efficiency optimization projects, and promote the transformation of energy-intensive industries to environmentally friendly development models by introducing advanced energy-saving equipment and innovative production processes. On the other hand, ETS internalizes the environmental costs of carbon emissions through carbon pricing. This mechanism forces firms to weigh the costs of carbon emissions against the benefits of investing in low-carbon technologies when making production decisions. As a result, firms are guided to reduce fossil fuel consumption while increasing investment in and adoption of clean energy, leading to a decline in carbon intensity. Furthermore, as a market-based instrument, ETS provides economic incentives for low-carbon technologies and clean energy while optimizing the allocation of emission reduction resources through the liquidity of carbon allowances. This method harmonizes financial gains with carbon-cutting objectives [51]. In summary, ETS establishes a policy foundation and market-driven forces for energy transition through its dual constraints on energy intensity and carbon intensity. Therefore, the following hypothesis is proposed:

Hypothesis 3:

ETS reduces energy intensity and carbon intensity by influencing the constraint behaviors of local governments, thereby improving regional energy transition.

4. Data and Methods

4.1. Projection Pursuit Model

Commonly used evaluation methods in academia include the AHP, the entropy method, and TOPSIS. However, these methods have inherent limitations. For example, AHP and TOPSIS are relatively subjective, while the entropy method can result in information loss. Consequently, the accuracy of evaluation outcomes is often questioned. To address these challenges, this study adopts a real-coded accelerating genetic algorithm-based projection pursuit model (RAGA-PPM). Compared with traditional methods like AHP (which is subjective) and entropy methods (which may ignore variable interdependence), PPM offers data-driven, non-parametric optimization that is particularly suited for high-dimensional, nonlinear, and non-normal data distribution. Although less common in social sciences, its robustness in pattern recognition makes it a powerful tool for composite index construction in energy transition research.

RAGA-PPM combines the PPM with the RAGA. The PPM, introduced by American scientist Kruskal in the 1970s, is a statistical method that integrates concepts from statistics, applied mathematics, and computer science. This approach is specifically designed to analyze and process high-dimensional observational data, particularly for non-linear and non-normal datasets [52]. The main steps for calculating projection values in PPM include standardizing the sample data, constructing a projection index function, optimizing the projection function, and calculating the projection values.

Given the complexity of optimizing the projection index function in PPM, RAGA is employed to enhance optimization speed and identify the optimal projection direction. This method can better overcome the Hamming Cliff problem of binary algorithms, with better optimization performance, concise coding, and high operability [53]. The main steps of RAGA for determining the optimal projection direction include: real-number encoding of optimization variables, initialization of the parent population, calculation of the fitness of the parent population, selection to generate the first offspring population, crossover operations, mutation operations, evolutionary iterations, and accelerated iterations to identify the optimal solution.

4.2. DID Model

The DID approach categorizes participants into treatment and control groups. By contrasting the disparities in the impacts of a particular event on these two groups, this method controls for unobservable factors that may confound causal relationships, thus identifying causality. In this study, the issuance of the Notice on Launching Pilot Projects for Carbon Emissions Trading is treated as a quasi-experiment, employing the DID approach to assess its influence on energy transition. The treatment group consists of six regions: Beijing, Shanghai, Tianjin, Guangdong, Hubei, and Chongqing, while the control group is composed of 24 provinces, except for Tibet. Considering the lag in policy implementation, the policy effects are assumed to begin in 2012. The specific model is presented as follows:

$$ET{L}_{it}=\alpha +\beta DI{D}_{it}+\gamma X_{it}+u_i+v_t+{\epsilon }_{it}$$

(1)

In the model, $i$ and $t$ represent regions and years. The variable ETL_it denotes the energy transition level. The variable DID_it represents the interaction of the time and regional dummy variables, where DID_it = 1 indicates that region i is a carbon trading pilot region in year t, and DID_it = 0 otherwise. X_it represents control variables, while $u_i$ and $v_t$ represent regional and temporal fixed effects. ${\epsilon }_{it}$ is the random error term.

4.3. Variables

4.3.1. Dependent Variable

The energy transition level (ETL) serves as the dependent variable in this analysis. By employing a multidimensional evaluation index system to quantify the ETL across 30 Chinese provinces between 2006 and 2021, the evaluation index system and calculation methods are provided in Section 3.1 and Section 4.1.

4.3.2. Independent Variable

The independent variable analyzed in this research is ETS (DID). It equals 1 if a province is a carbon trading pilot region in 2012 or later; otherwise, it equals 0.

4.3.3. Control Variables

Based on existing research, this study selects economic development level (ve), technological innovation (giv), urbanization level (urb), openness to the outside world (fdi), human capital (hc), and industrial structure (is) as control variables [9,54,55,56]. ve is gauged using per capita GDP, giv by tallying green patent applications, urb is determined by the ratio of the urban populace to the overall population, fdi by the yearly foreign direct investments, hc reflects the average count of higher education students for every 100,000 individuals, and is gauges the share of the secondary sector in the GDP.

4.3.4. Mediator Variable

This study selects local government behavior as the mediator variable. Based on theoretical analysis, local government behavior is categorized into incentive-based and constraint-based actions. Incentive-based behavior refers to government support for regional technological innovation (R&D). Following Dong et al. (2022), R&D expenditure intensity serves as a measure of incentive-based behavior [57]. Constraint-based behavior reflects the intensity of environmental regulation by local governments, which includes energy intensity constraints (EI) and carbon intensity constraints (CI). Higher values of these indicators indicate weaker environmental regulation, whereas lower values represent stronger regulation intensity.

4.4. Data Description

This study analyzes panel data from 30 provinces in mainland China spanning the period 2006 to 2021 to evaluate the impact of ETS on energy transition. The data were obtained from the China Energy Statistical Yearbook, China Statistical Yearbook, China Science and Technology Statistical Yearbook, provincial statistical yearbooks and bulletins, and the National Intellectual Property Administration. To mitigate the effects of non-stationarity in macro-level data, all variables were logarithmically transformed. A summary of the variables’ descriptive statistics is provided in

Table 2. Descriptive statistics of variables.

Variable Names	Abbreviation	Observations	Mean	S.d.	Min	Max
carbon emissions trading	DID	480	0.13	0.33	0	1
energy transition level	ETL	480	0.45	0.30	−0.44	1.23
government’s technological support	R&D	480	1.60	1.11	1.11	6.53
energy intensity constraints	EI	480	0.92	0.56	0.17	3.90
carbon intensity constraints	CI	480	2.81	1.98	1.98	12.13
economic development level	ve	480	10.61	0.62	8.66	12.12
urbanization level	urb	480	4.00	0.24	3.31	4.50
human capital	hc	480	7.79	0.35	6.74	8.82
industrial structure	is	480	3.77	0.24	2.76	4.12
technological innovation	giv	480	7.39	1.64	2.40	10.94
openness to the outside world	fdi	480	15.50	1.49	11.22	19.43

.

5. Results and Discussion

5.1. Energy Transition Level Measurement and Spatial-Temporal Distribution

Due to space limitations, the detailed process of RAGA-PPM is not described here. Based on the evaluation results of energy transition, distribution maps of China’s energy transition levels for 2006, 2011, 2016, and 2021 were generated (Figure 1). A clear spatial heterogeneity is evident. Provinces such as Beijing, Shanghai, and Jiangsu consistently rank at the top in terms of energy transition levels. These regions benefit from advanced economic development, comprehensive low-carbon policy frameworks, and strong technological and financial capacities. For instance, Beijing has implemented strict carbon intensity control policies and has prioritized green R&D investment, supported by its status as a national policy pilot region. Similarly, Shanghai’s local government has integrated carbon trading into broader climate governance strategies, including green finance and emission intensity benchmarking across industries.

In contrast, provinces such as Inner Mongolia, Ningxia, and Shanxi remain at the lower end of the energy transition index. These areas are heavily reliant on coal-based energy systems and resource-extractive industries, which pose significant obstacles to structural transformation. Despite national mandates, local implementation capacity is often constrained by fiscal dependence on high-emission sectors and limited access to clean energy technologies. For example, Inner Mongolia’s economy still relies significantly on coal mining and heavy industry, making the shift to renewables both economically and politically sensitive.

Furthermore, the trends in average energy transition levels across regions during the sample period were analyzed, as illustrated in Figure 2. Based on socioeconomic development, China’s 30 provinces are categorized into eastern, central, western, and northeastern regions. The findings indicate that the eastern region has achieved a more advanced energy transition compared with the central, western, and northeastern regions. This can be attributed to the eastern region’s advanced economic development., a well-established clean energy equipment manufacturing industry, stronger technological innovation capabilities, and a solid industrial foundation, all of which provide robust support for energy transition. Moreover, as economic development advances, the growing public demand for improved environmental quality further enhances the willingness to pursue energy transition in the eastern region.

The Qinling–Huaihe Line is a significant geographical boundary in China, dividing the country’s 30 provinces into southern and northern regions. As indicated in the analysis, during the sample period, the south persistently demonstrated a greater level of energy transition compared with the north. The provinces with the lowest energy transition levels—Ningxia, Shanxi, and Inner Mongolia—are all located in the northern region. Although Beijing has the highest energy transition level, this can likely be attributed to a combination of political, economic, and cultural factors, given its status as the capital of China. Furthermore, southern provinces such as Shanghai and Jiangsu, which demonstrate relatively high energy transition levels, further validate the findings of this study. The lower energy transition levels in the northern region may be explained by the substantial energy consumption required for winter heating, which relies heavily on traditional fossil fuels due to the region’s limited clean energy resources [58].

5.2. DID Regression Results

Using Equation (1), this section delves into a quantitative assessment of the impact of ETS implementation on energy transition.

Table 3. DID regression results.

Variable	(1)	(2)
	ETL	ETL
DID	0.0105	0.0483 ***
	(0.0144)	(0.0154)
giv		−0.0113
		(0.0176)
ve		0.191 ***
		(0.0448)
urb		0.255 **
		(0.101)
hc		0.0866 *
		(0.0524)
is		−0.410 ***
		(0.0634)
fdi		−0.0446 ***
		(0.0107)
Observations	480	480
R-squared	0.946	0.955

Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

presents the estimated results. In Column (1), the results are presented without any control variables, while Column (2) incorporates them. As shown in Table 3, after incorporating the control variables, the DID regression coefficient is 0.0483, which is significant at the 1% level. This indicates that ETS implementation has enhanced the energy transition level in the pilot provinces, with an improvement of 4.83%, thereby supporting Hypothesis 1. Although the percentage may appear moderate, it is economically meaningful given the complexity of structural energy transformation. According to China’s 14th Five-Year Plan, achieving measurable progress in energy diversification and carbon reduction within a decade is a critical milestone. A 4.83% increase suggests that ETS can serve as a catalyst for accelerating long-term structural changes, especially when scaled nationally.

5.3. Robustness Tests

5.3.1. Parallel Trend Test

Drawing inspiration from the methodology outlined by Deschênes et al. (2017), this research leverages the event study approach to perform a parallel trend analysis [59]. As illustrated in Figure 3, the dashed line marks the point at which the ETS was rolled out. Notably, the energy transition in both the treatment and control groups displayed comparable trends before the ETS’s implementation, indicating that the parallel trend assumption holds.

5.3.2. Placebo Test

To address potential bias from unobserved variables in the initial regression analysis, a placebo test was performed using the approach of Cai et al. (2016) [60]. As illustrated in Figure 4, the regression coefficients exhibited a notable divergence from the true values, and the kernel density distribution of the observed values was clustered around 0. Therefore, the empirical results are robust.

5.3.3. Counterfactual Test

When applying the DID method, ensuring the comparability of the treatment and control groups is essential; otherwise, causal identification may be biased. To address this issue, counterfactual tests were performed by constructing fictitious policy dummy variables, where the implementation of the ETS was hypothetically advanced by 2 and 4 years. These fictitious policy variables were then incorporated into Equation (1) for testing.

Table 4. Counterfactual test.

Variable	(1)	(2)
	ETL	ETL
DID2/DID4	0.0292	0.00254
	(0.0179)	(0.0240)
giv	−0.0119	−0.0120
	(0.0177)	(0.0177)
ve	0.206 ***	0.211 ***
	(0.0446)	(0.0448)
urb	0.201 **	0.170 *
	(0.0979)	(0.0988)
hc	0.0811	0.0718
	(0.0532)	(0.0525)
is	−0.413 ***	−0.412 ***
	(0.0642)	(0.0645)
fdi	−0.0450 ***	−0.0443 ***
	(0.0108)	(0.0109)
Observations	480	480
R-squared	0.955	0.955

Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

displays the results. The analysis reveals that, assuming that the policy rollout was pushed forward by 2 and 4 years, the estimated coefficients of DID2 and DID4 were both statistically insignificant. This demonstrates that the ETS has enhanced energy transition levels in the regions.

5.3.4. PSM-DID Estimation

To address potential sample selection bias, we utilized propensity score matching (PSM). In logit regression analysis, we incorporated control variables as covariates to estimate the propensity scores for every region. The control group was paired with the treatment group in a 1:2 ratio through nearest-neighbor matching. As indicated in

Table 5. PSM-DID estimation.

Variable	ETL
DID	0.151 **
	(0.0598)
giv	0.188 ***
	(0.0198)
ve	0.0655
	(0.0890)
urb	−0.501 **
	(0.254)
hc	−0.0916
	(0.101)
is	−0.269 ***
	(0.0774)
fdi	0.0349 **
	(0.0154)
Constant	1.756
	(1.140)
Province FE	YES
Year FE	YES
Observations	190
R-squared	0.684

Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05.

, the coefficient of DID stood at 0.151 and was statistically significant at 5%. This further shows that the regression results are robust.

5.4. Heterogeneity Analysis

5.4.1. Level of Marketization

The core concept of the ETS is to treat carbon emission rights as tradable commodities among enterprises. The successful operation of the ETS hinges on the trading mechanism’s efficiency, where marketization plays a pivotal role. Marketization reflects resource mobility between low- and high-efficiency areas [61]. As a result, differences in the levels of marketization across regions could result in notable discrepancies in how the ETS influences the energy transition. Against this backdrop, this study conducts a heterogeneity analysis based on the regional marketization level. The marketization level is measured using the marketization index developed by Fan et al. (2003) [62].

The findings of the heterogeneity analysis, detailed in Column (1) of

Table 6. Heterogeneity regression results.

Variable	ETL
	(1)	(2)
marketization	0.0894 *
	(0.0525)
M1	0.153 **
	(0.0697)
logpatent		−0.0150
		(0.0177)
M2		0.0250 **
		(0.0119)
Control	YES	YES
Province FE	YES	YES
Year FE	YES	YES
Observations	360	360
R-squared	0.768	0.752

Robust standard errors in parentheses. ** p < 0.05, * p < 0.1.

, reveal how different levels of marketization influence the results. The analysis reveals a coefficient of 0.153, which is statistically significant at 5%. This suggests that higher levels of marketization amplify the effectiveness of ETS, likely due to the more efficient allocation of emission rights, stronger enforcement capacity, and better-developed institutional infrastructure for carbon trading. In such environments, the price signal of carbon becomes more transparent and actionable for firms.

5.4.2. Regional Innovation Capacity

Enterprises, as key market participants, may reallocate innovation resources in response to environmental regulatory policies to meet the requirements of energy conservation [6]. Consequently, regional differences in innovation capacity may lead to heterogeneous effects of the ETS on energy transition. This study incorporates an interaction term between innovation capacity and ETS (M2) into the regression analysis. According to Column (2) in Table 6, the regression coefficient stands at 0.250 and is statistically significant at 5%. This indicates that the impact of the energy transition associated with the ETS varies significantly across regions. Regions with strong innovation capacity are better positioned to absorb technological shocks, rapidly adopt cleaner technologies, and engage in green R&D, which enhances the marginal utility of ETS policies.

5.4.3. Quantile Regression

The quantile regression model reveals the extent of variability in y attributable to x, as well as detailed information on the conditional distribution of $y|x$. In this study, energy transition is used as y, while the implementation of the ETS serves as x to construct the following quantile regression model:

$$Q_T\left[Y|X\right]={\beta }_{0T}+{\beta }_{1T}{D}_i+{\epsilon }_T$$

(2)

In Equation (2), $Q_T\left[Y|X\right]$ represents the outcome variable, which indicates the energy transition level at the T quantile for a given region. $D_i$ is the key explanatory variable, representing whether the region has implemented the ETS. ${\beta }_{0T}$ denotes the constant term, ${\beta }_{1T}$ is the coefficient of $D_i$ estimated at the T quantile, and ${\epsilon }_T$ represents the random disturbance term.

This study employs panel data quantile regression at the 5th, 25th, 50th, 75th, and 90th percentiles as quantile points.

Table 7. Quantile regression results.

Variable		ETL
	5%	25%	50%	75%	90%
DID	0.653 ***	0.373 ***	0.351 ***	0.405 ***	0.461 ***
	(0.0461)	(0.0316)	(0.0444)	(0.0666)	(0.0496)
Constant	−0.116 ***	0.280 ***	0.411 ***	0.556 ***	0.721 ***
	(0.0328)	(0.0216)	(0.0136)	(0.0185)	(0.0241)
Observations	480	480	480	480	480

Robust standard errors in parentheses. *** p < 0.01.

displays quantile regression outcomes. The implementation of the ETS exerts a significant impact on energy transition levels across all quantiles. Notably, the effect is stronger in lower quantiles (e.g., 5%), indicating that the ETS may offer greater marginal benefits in regions with initially weaker energy transition performance. These regions often suffer from outdated industrial structures, high coal dependency, and limited clean energy infrastructure, making them more responsive to ETS-induced incentives and constraints. In contrast, higher-quantile regions may already possess mature energy systems and institutional capacity, thus exhibiting relatively smaller incremental gains.

6. Mechanism Analysis: The Role of Local Government Behavior

The findings above confirm that the ETS promotes regional energy transition. However, through which channels does the ETS exert its effects? Theoretical analysis suggests that local governments enhance regional energy transition through both incentive-based and constraint-based mechanisms when implementing the ETS. To assess the impact of these two mechanisms—government incentives and constraints—on regional energy transition, this research employs the methodology of Wen et al. (2014) and formulates the subsequent model [63].

$$ET{L}_{it}=\alpha +\beta DI{D}_{it}+\gamma X_{it}+u_i+v_t+{\epsilon }_{it}$$

(3)

$$M_{it}={\alpha }_0+{\beta }_{0}DI{D}_{it}+{\gamma }_0{X}_{it}+u_i+v_t+{\epsilon }_{it}$$

(4)

$$ET{L}_{it}={\alpha }_1+{\beta }_{1}DI{D}_{it}+{\beta }_2{M}_{it}+{\gamma }_1{X}_{it}+u_i+v_t+{\epsilon }_{it}$$

(5)

where M represents local government behavior, which includes government support for technology (R&D), energy intensity constraints (EI), and carbon intensity constraints (CI). The meanings of other symbols in this equation are consistent with those in Equation (1).

The regression analysis presented in Column (1) of

Table 8. Results of the mechanism analysis.

	Incentive Behaviors		Constraint Behaviors
	(1)	(2)	(3)	(4)	(5)	(6)
Variable	R&D	ETL	EI	ETL	CI	ETL
DID	0.315 ***	0.0223	−0.0354	0.0438 ***	−0.195 **	0.0359 **
	(0.0725)	(0.0152)	(0.0269)	(0.0150)	(0.0819)	(0.0150)
R&D		0.0825 ***
		(0.0136)
EI				−0.127 ***
				(0.0325)
CI						−0.0639 ***
						(0.0100)
Control	YES	YES	YES	YES	YES	YES
Province FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
Observations	480	480	480	480	480	480
R-squared	0.966	0.958	0.952	0.958	0.960	0.962

Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05.

indicates that the introduction of the ETS has significantly boosted R&D. The impact coefficient is 0.0315, and this result is statistically significant at the 1% level. This finding suggests that the ETS increases government R&D funding by influencing local government incentive behaviors. The findings in Column (2) show that the regression coefficient for R&D’s impact on energy transition is 0.0825, significant at the 1% level. This demonstrates that local government incentive behaviors promote regional energy transition by increasing R&D funding. Therefore, Hypothesis 2 is confirmed.

Columns (3)–(6) present the findings regarding government constraint behaviors. Column (3) reveals that the ETS fails to produce a statistically meaningful influence on local government constraints tied to energy intensity. On the other hand, Columns (5) and (6) suggest that the ETS does play a significant role in curbing regional carbon intensity. This constraint, in turn, compels regions to enhance their energy transition levels. Therefore, Hypothesis 3 is partially confirmed.

7. Conclusions and Policy Implications

This study examines the impact of China’s carbon emissions trading scheme (ETS) on regional energy transition using panel data from 30 provinces over the period of 2006–2021. By constructing a multi-dimensional energy transition index and applying a difference-in-differences (DID) approach, we find that the ETS significantly promotes regional energy transition, with pilot areas achieving an average improvement of 4.83%. Further analysis reveals that the ETS primarily promotes energy transition by influencing local government behaviors. Incentive behaviors are realized through enhanced financial and technological support, while the constraining effects are manifested in restrictions on carbon emissions intensity. The effect is more pronounced in regions with stronger marketization and higher innovation capacity, and is particularly significant in provinces with initially low levels of energy transition, indicating a progressive and convergence-enhancing impact.

These findings suggest that the carbon market can serve as an effective tool not only for emissions control, but also for advancing long-term structural transformation in energy systems. To enhance the effectiveness of the ETS, it is essential to gradually improve the allowance allocation mechanism. In the short term, benchmark-based free allocation should be tightened and dynamically adjusted to reflect technological progress. Over time, a transition toward a mixed system that incorporates auctions—particularly in sectors with mature monitoring and verification systems—will allow carbon pricing to better reflect marginal abatement costs. Revenues from auctions can be directed toward green R&D and capacity building in less-developed regions to ensure a just and equitable transition.

Additionally, differential policy strategies should be adopted to account for regional heterogeneity. In innovation-intensive and market-oriented regions, more flexible carbon trading instruments and financial derivatives can be introduced to improve market liquidity and efficiency. In contrast, regions with weaker institutional capacity and industrial dependence on fossil fuels will require stronger regulatory support, capacity-building initiatives, and transitional safeguards. Tailoring ETS implementation to local conditions will be crucial for maximizing its transformative potential and achieving China’s dual carbon goals in an inclusive and balanced manner.

While this study offers robust empirical evidence, several limitations should be acknowledged: (1) The construction of the energy transition index, though data-driven, still depends on available indicators and may not capture all qualitative factors such as policy enforcement. Future research could incorporate firm-level data or cross-country comparisons to enrich the analysis. (2) The analysis does not consider potential spatial spillover effects and carbon leakage, future research will incorporate spatial econometric models and cross-regional emission dynamics to further explore these issues.