To Cooperate Proactively or Pay Fines? Unpacking the Dual Effects of Government Intervention and Market Incentives on Carbon Emissions Intensity in Power Enterprises

Abstract

This study identified the impact patterns of carbon price fluctuations and excess penalty mechanisms on power companies’ carbon emissions from the dual perspectives of market incentives and government intervention to address the issues of insufficient incentives for power companies to reduce emissions and environmental regulation mechanisms for emission reduction. Three main conclusions are revealed by the study: First, carbon emissions are significantly suppressed by the Environmental Protection Law of the People’s Republic of China (EPLPRC), with an inverse U-shaped link between carbon allowance pricing (CAP) and carbon emissions intensity (CEI). Moreover, the two exhibit synergistic effects. Second, two important transmission mechanisms are unit material cost (UMC) and green technological innovation (GHI). Third, there is significant variation in the results of emission reduction. While the inverted U-shaped effect of carbon allowance prices is most noticeable among medium-to-large non-state-owned businesses and in areas with rapid increase in thermal power capacity, EPLPRC works better for small non-state-owned businesses and large state-owned businesses. We also note that in areas where the expansion of thermal power generation is slower, dual environmental policies show significant synergistic impacts. These results offer direction for developing distinct carbon reduction strategies and promoting the power industry’s low-carbon transition.

1. Introduction

The rapid growth of the global economy has led to severe environmental pollution, particularly excessive carbon emissions. To address climate change and effectively reduce carbon emissions, countries have introduced various environmental regulations. According to the International Energy Agency (IEA) and the latest research, global greenhouse gas emissions are projected to reach 37.8 billion tons in 2024, an increase of 0.8% compared to 2023. Over 75% of these emissions stem from the combustion of fossil fuels. Electric power companies, heavily reliant on fossil fuels, have become primary targets for emissions reduction. Scholars Yuan et al. (2025) [1] note that due to market failures and the profit-seeking nature of capital, these enterprises lack both the incentive and the capacity to reduce emissions. Consequently, driving high-emission power companies to cut carbon has become an urgent priority.

From a government intervention perspective, companies often increase green investments to reduce carbon emissions when facing stringent environmental regulations, thereby avoiding penalties [2]. Environmental pollution generates significant negative externalities, making government intervention crucial for correcting market failures. Effective governance is a key prerequisite for successful intervention [3]. In the 1970s, developed nations tackled pollution through mandatory regulations, exemplified by the U.S. Environmental Protection Agency (EPA) and the Clean Air Act—classic models of command-and-control environmental regulation (CCER) [4,5]. However, the effectiveness of CCER in reducing carbon emissions has been questioned due to high costs, lack of flexibility and increasingly complex economic environments and ecological challenges.

Meanwhile, market-based incentive environmental regulations are gaining increasing favor. Typical market-based incentive environmental regulations include the European Union Emissions Trading System (EU ETS), the U.S. sulfur dioxide (SO₂) emissions trading program, and the Nordic carbon tax policy. From a market incentive perspective, when enterprises receive benefits such as emission reduction subsidies or free emission allowances, their motivation to reduce emissions becomes stronger.

Currently, many developing countries have begun implementing environmental regulations relatively late, initially relying heavily on the experiences of developed nations. They primarily adopted CCER to reduce pollution, but implementation outcomes have often been unsatisfactory due to insufficient technology, funding, and management capacity. In recent years, some developing countries have begun adopting market-based environmental incentive mechanisms. For instance, China formally launched its national carbon emissions trading market in 2021, allowing enterprises to meet regulatory requirements by purchasing carbon allowances or directly paying fines [6,7,8]. India’s renewable energy subsidy policy has also effectively reduced carbon emissions levels.

Currently, most researchers analyze carbon reduction mechanisms by focusing on individual environmental policies. For instance, Lin et al. (2023) adopted a government intervention perspective, using a difference-in-differences approach to demonstrate that the implementation of Environmental Protection Law of the People’s Republic of China (EPLPRC) significantly reduced carbon emission intensity in industrial sectors [9]. The impact was particularly pronounced in industries characterized by higher market competition and larger scale. The EU Emissions Trading Scheme (EU ETS) significantly impacts the power and cement industries by influencing unit material costs, reflecting the costs of complying with emission caps or other parallel renewable energy incentive programs [10].

Some studies focus on a single scale of formal and informal types, examining the carbon reduction effects of different environmental regulations from macro levels and industry levels. It is widely acknowledged that formal environmental regulations prove more effective in east of China, large cities, and high-carbon-emission cities, while informal environmental regulations are more effective in underdeveloped small and medium-sized cities and western regions [11]. Shi et al. (2024) [12] confirmed that China’s carbon market effectively constrained carbon emissions in the power sector.

The primary aim of this study is to dissect the complex interplay between government intervention and market incentives in shaping corporate CEI within the power sector. Specifically, we seek to understand how environmental regulations, through both command-and-control measures and market-based incentives, influence power firms’ emission reduction behaviors. We utilize data from Chinese power firms spanning 2013 to 2021 to empirically examine the conditional nature of emission reduction effects. The overarching goal is to provide actionable insights for policymakers to design effective environmental regulations that promote sustainable development and carbon market efficiency. Examining perspectives from both government intervention and market incentives, we specifically investigated how power enterprises respond to increasingly stringent market-based regulation: whether they opt for “low-profile compliance” by paying fines or proactively invest in green innovation technologies and participate in carbon quota trading to reduce emissions.

The marginal contributions of the study are primarily reflected in three aspects. First, we constructed a dynamic theoretical model. Adopting a dual perspective of government intervention and market incentives, we elucidated the dynamic interaction mechanism between environmental regulations and firms’ strategic responses to carbon emissions. Second, compared to previous analyses at the provincial or sectoral level, our examination of firm-level data enables a more granular assessment of environmental regulations’ carbon reduction effects. We demonstrate that the emission-reducing impact of environmental regulations is conditional, closely tied to carbon price fluctuations and influenced by the concurrent implementation of the EPLPRC under government intervention. Third, our heterogeneity analysis—grouped by the expansion rate of thermal power supply capacity—confirms that EPLPRC proves more effective in regions with faster energy transitions, while carbon markets yield greater impact in areas with rapid thermal power supply expansion. This provides concrete and actionable insights for designing and implementing different types of environmental policy regulations.

The remainder of this paper is organized as follows: Section 2 reviews the literature, constructs the theoretical model, and proposes hypotheses; Section 3 outlines the empirical research design; Section 4 presents results analysis and extended discussion; Section 5 offers policy recommendations and concludes.

2. Theoretical Background and Hypothesis Development

2.1. Theoretical Background

Based on institutional theory, corporate production and management practices are shaped not only by internal resource endowments and strategic considerations but more significantly by external institutional environments [13,14]. Within this framework, governments—as key designers of institutional environments—intervene in corporate operations with the objective of maintaining a dynamic equilibrium between low-carbon economic development and ecological conservation [15].

A core unresolved issue in academia remains: Can environmental regulations effectively reduce carbon emissions? This debate centers on two opposing perspectives: the “green paradox” and the “Porter hypothesis.” Early theoretical research, exemplified by the Porter hypothesis [16], argues that well-designed environmental regulations can trigger innovation compensation effects that offset or even exceed compliance costs, thereby reducing carbon emissions. Porter and Van der Linden (1995) contend that well-structured regulations incentivize firms to pursue technological innovation [16,17]. This drives the adoption of cleaner production models, ultimately achieving energy conservation and emission reduction [18,19]. Subsequent empirical studies broadly support this view. For instance, Cairns (2014) [20] found environmental regulations significantly reduce carbon emissions, a conclusion unaffected by the natural and technological characteristics of oil production. Yin et al. (2015) [21] observed that environmental regulations can prompt regional relocation of high-carbon industries, indirectly achieving emission reduction targets. Tang et al. (2025) [22] demonstrated that market-oriented environmental regulations effectively curb emissions of carbon dioxide, sulfur dioxide, and particulate matter.

However, real-world developments revealed an unexpected trend: despite widespread implementation of environmental regulations, particularly following the Kyoto Protocol, global carbon emissions not only failed to decline but increased. This phenomenon prompted scholars to reassess the impact of environmental regulations on carbon emissions, leading to the formulation of the Green Paradox hypothesis (Sinn, 2008) [23].

The green paradox posits that when firms anticipate stricter future environmental regulations, they may accelerate the extraction and consumption of existing fossil fuel resources, leading to increased current energy use and higher carbon emissions [24,25,26]. The theory suggests that companies within the fossil fuel supply chain face reduced expected future profits when confronted with tighter environmental regulations [27,28]. To mitigate potential losses, these companies may increase fossil fuel extraction in the short term, leading to market oversupply and significant price declines [29]. Such price drops further stimulate demand for fossil fuels, causing a temporary increase in carbon intensity [30]. This phenomenon is particularly pronounced during the early stages of environmental regulation.

2.2. Theoretical Model and Research Hypothesis

The model explores the dynamic complexity of the relationship between environmental regulations and carbon emissions. It provides a mechanism-based explanation for the sustainable transformation of the power sector.

Carbon emissions, an inevitable byproduct of power generation, directly correlate with a company’s production capacity. Generally, higher electricity generation leads to increased carbon emissions. Based on this relationship, we establish the benchmark emission function:$e=\beta ·q$, where $e$ represents carbon emissions; $q$ denotes electricity generation, reflecting the company’s production capacity; $\beta$ is the carbon emissions intensity per unit output, indicating the production technology efficiency. This equation shows that carbon emissions increase linearly with electricity generation. A higher value of $\beta$ indicates greater carbon emissions per unit of electricity produced. Under the carbon market mechanism, enterprises initially receive government-allocated free carbon allowance ($E$). When actual carbon emissions ($e$) exceed the allowance ($E$), enterprises must address the shortfall by purchasing additional carbon allowances through market transactions or incurring penalties for non-compliance.

First, consider the enterprise’s carbon allowance purchasing behavior. Since enterprises may opt for partial compliance with regulations. The actual carbon emissions purchased represent a proportion $\rho$ ($0<\rho <1$) of $\left(e-E\right)$. $\rho$ quantifies the enterprise’s sensitivity to carbon price fluctuations. The actual expenditure for purchasing carbon allowances is $G=\kappa ·\rho ·\left(e-E\right)$. Here, $\kappa$ represents the unit carbon price, $\rho$ denotes the firm’s actual purchase ratio, and $\rho$ is directly influenced by the carbon price $\kappa$. The two exhibit a negative correlation, with the specific functional form being $\rho =c_i·\kappa \left(c_i<0\right)$. In this context, $c_i$ is the carbon price sensitivity coefficient, where a negative value indicates a decrease as the price rises.

Second, for non-compliant portions, the government will impose administrative penalties. The enterprise’s breach-of-contract loss is $L=F·\left(1-\rho \right)·\left(e-E\right)$. Here, $F$ represents the unit penalty standard for excess emissions, reflecting the strictness of regulation. Combined with the relationship between production capacity and carbon emissions, these decision variables collectively form the enterprise’s cost constraint.

The firm’s comprehensive utility, $U_i$ is influenced by the revenue from carbon emissions activities, the cost of purchasing carbon allowances, and the penalty losses from breach. Its functional form is defined as:

$$U_i=\left(a-{\displaystyle \frac{b·e}{\beta }}\right)·{\displaystyle \frac{e}{\beta }}-\kappa ·\rho ·\left(e-E\right)-\mathrm{F}·\left(1-\rho \right)·\left(e-E\right)$$

(1)

In the equation, coefficients $a$ and $b$ are both greater than zero, representing the base revenue from production and the marginal negative utility of carbon emissions, respectively, reflecting the degree of internalization of environmental costs. To maximize utility, the following conditions must be satisfied.

$${\displaystyle \frac{{\partial U}_i}{\partial e}}={\displaystyle \frac{a}{\beta }}-{\displaystyle \frac{2be}{{\beta }^2}}-c_i·{\kappa }^2-F·\left(1-c_i·\kappa \right)=0$$

(2)

$$e={\displaystyle \frac{{\beta }^2}{2b}}\left({\displaystyle \frac{a}{\beta }}-c_i{·\kappa }^2-F+F·c_i·\kappa \right)$$

(3)

Equation (3) represents the optimal carbon emissions decisions for a firm under given decision variables $F$ and $c_i$. By deriving the firm’s optimal decisions and further taking partial derivatives, we analyze the marginal impact of key variables on the optimal carbon emissions levels.

$${\displaystyle \frac{\partial e}{\partial F}}=-{\displaystyle \frac{{\beta }^2}{2b}}\left(1-c_i·\kappa \right)$$

(4)

$${\displaystyle \frac{\partial e}{\partial c_i}}={\displaystyle \frac{{\beta }^2}{2b}}\left(-{\kappa }^2+F·\kappa \right)$$

(5)

From Equation (4), we observe that ${\displaystyle \frac{\partial e}{\partial F}}<0$. That is, the higher the excess emission penalty threshold, the lower the optimal carbon emissions level $e$ for enterprises, indicating that the penalty mechanism exerts a significant suppression effect on carbon emissions. Based on this, we propose the following hypothesis to be tested.

H1. 

From the perspective of government intervention, fines for excess emissions have significantly curbed corporate carbon intensity.

The carbon emissions trading system aims to limit corporate pollutant emissions, fundamentally addressing market failures caused by environmental externalities. However, this system may impose additional cost pressures on enterprises exceeding emission quotas. Such cost pressures could weaken corporate competitiveness, potentially leading to increased rather than reduced carbon emissions [31]. China’s current carbon market remains imperfect, with carbon prices lower than those in the EU market and prone to significant volatility, resulting in considerable uncertainty in carbon price signals. Consequently, we propose that environmental regulations influence corporate carbon emissions through dual mechanisms of “cost constraints” and “revenue drivers.” The marginal effects of these two mechanisms dynamically shift with regulatory intensity, ultimately forming a nonlinear causal relationship.

Regarding the relationship between environmental regulations and carbon emissions, Guo and Chen (2018) [27] observe that while mainstream views assume a linear impact, practice shows both forced emission reductions and the green paradox coexist. This indicates the relationship is more likely nonlinear. We argue that the relationship fundamentally reflects a dynamic trade-off between varying regulatory intensities and the costs of low-carbon transitions. In early regulatory stages—when carbon prices are low—the cost of voluntary emission reductions far exceeds penalties and transaction costs. At this point, firms lack incentives for proactive carbon cuts. However, as regulations mature and strengthen, technological advancements and heightened environmental awareness drive emissions downward.

Fluctuations in carbon prices further complicate corporate carbon emissions decisions. These manifest in two scenarios: low-price and high-price conditions. When the carbon price ($\kappa$) is low, firms typically opt to directly purchase allowances. This approach proves cheaper than investing in emissions reductions, leading them to avoid proactive emission cuts.

The relationship between corporate compliance behavior and carbon pricing is not a simple linear correlation, but rather a dynamic equilibrium influenced by multiple factors. Rising carbon prices may incentivize companies to reduce emissions through economic incentives, but the specific outcomes depend on factors such as corporate heterogeneity, technological feasibility, and policy design.

Firms’ sensitivity to the CAP is modulated by market conditions, such as carbon price levels and penalty intensity), resulting in phased variations in their compliance behavior. The sign of ${\displaystyle \frac{\partial e}{\partial c_i}}$ in Equation (5) depends on the relative magnitude of the $F$ and $\kappa$. It indicates that firms’ response to carbon prices is not a simple linear positive correlation but rather depends on market conditions. Based on this finding, we propose the following hypothesis for testing.

H2. 

The volatility of carbon prices exhibits an inverted U-shaped nonlinear relationship with power companies’ carbon emissions.

3. Research Design

3.1. Data

We constructed our research sample using data from Chinese listed power companies between 2013 and 2021. After removing missing values, the final sample comprised 2140 observations from 289 power companies. Data primarily originated from the National Bureau of Statistics of China, the China Stock Market and Accounting Research Database (CSMAR), and the China National Research Data Service (CNRDS) database. To ensure data quality, we applied the following screening criteria to the raw data: Excluded listed companies with severe financial data deficiencies; Excluded listed companies with assets below liabilities; Excluded samples marked as ST or *ST during the year; Excluded financial institutions such as banks, insurance companies, securities firms, and trust companies. To minimize the interference of extreme values on the study, the data underwent trimming.

3.2. Variables

3.2.1. Dependent Variable

Amid efforts to advance low-carbon transformation and development, carbon emissions intensity has become a key research focus. The dependent variable in this study is CO₂ emissions per unit of output to express carbon. Carbon emissions data is sourced from ESG reports, social responsibility reports, and annual reports.

3.2.2. Independent Variables

The core explanatory variables in this study are excess emissions fines (EEF) and carbon allowance prices (CAP). From the perspective of government intervention, EPLPRC, implemented in 2015, established stringent measures such as daily consecutive penalties and production restrictions or shutdowns. We define EEF as the original penalty amount (in ten thousand yuan) disclosed for violations of EPLPRC during the reporting period. If a company didn’t receive penalties for such violations in the year, EEF is assigned a value of 0. Data sources include annual reports of listed companies, corporate social responsibility reports, and the environmental information disclosure section of the CSMAR database.

From a market incentive perspective, the essence of carbon markets lies in internalizing the external costs of carbon emissions through price signals. The level of carbon prices directly influences corporate carbon emission intensity. Due to differing daily benchmark price calculation rules across pilot carbon markets—for instance, Beijing and Tianjin use the daily average transaction price as the benchmark, while Shanghai and Hubei employ the weighted average of the last five daily transactions as the closing price. To mitigate interference from a few abnormal transaction prices and ensure the measured CAP is applicable to both large-scale pilot carbon markets with ample data like Guangdong and Hubei, as well as moderately active pilot regions like Chongqing, we aggregate daily valid transaction data from all pilot carbon markets throughout the year. The annual CAP is then calculated using the formula:

$${CAP}_{Year}={\displaystyle \raisebox{1ex}{$\sum \left({CAP}_{Daily}·{Volume}_{Daily}\right)$}\!\left/ \!\raisebox{-1ex}{${Volume}_{Year}$}\right.}$$

(6)

3.2.3. Control Variables

To avoid interference from relevant factors and ensure the reliability of research conclusions, we selected the following control variables. Firm size (lnsize) influences market power, bargaining power, risk diversification, and other factors. The measurement method employs the natural logarithm of total assets within the fiscal year. Leverage ratio (Lev) measures a firm’s financial leverage level. A high leverage ratio may indicate heightened financial risk, while a low ratio may suggest greater financial stability and debt-servicing capacity. Return on Equity (ROE) primarily reflects a firm’s ability to generate income and profits from market activities. Return on Assets (ROA) measures the contribution of resource utilization efficiency to net profit. firms with strong profitability typically demonstrate greater environmental awareness, reflected in this metric as the ratio of net profit to total assets. Earnings Per Share (EPS) represents a company’s ability to create value for shareholders. Incorporating EPS as a control variable can isolate the potential impact of financial performance on carbon intensity, as financially sound enterprises may allocate more resources to environmental management. Tobin’s Q ratio (Tobin’s Q) mainly refers to the ability of an enterprise to obtain income and profits from the market.

To eliminate potential confounding factors such as “regions with high carbon prices having higher renewable energy shares,” “lower electricity demand in regions with high carbon prices,” and “rising coal prices when carbon prices are high,” we incorporate three variables: the proportion of thermal power generation capacity (TPGC), regional electricity demand (RED) and coal prices (CP), as control variables. This approach enables a clearer observation of the relationship between environmental regulations and carbon emissions from power companies.

3.3. Descriptive Statistic

The descriptive statistical results are shown in

Table 1. Descriptive analyses.

Variables	N	Mean	SD	Min	p25	p50	p75	Max
CEI	2140	0.708	0.634	0.059	0.098	0.662	1.127	2.384
EEF	1885	0.507	0.534	0.004	0.162	0.325	0.645	3.319
CAP	2140	29.830	14.870	10.310	17.930	28.160	39.410	78.890
ROE	2140	0.049	0.136	−0.760	0.024	0.062	0.102	0.348
Lev	2140	0.497	0.190	0.111	0.359	0.499	0.639	0.952
TobinQ	2140	1.778	1.132	0.837	1.135	1.440	1.985	8.484
EPS	2140	0.306	0.492	−1.280	0.070	0.235	0.490	2.205
ROA	2140	0.032	0.053	−0.194	0.011	0.030	0.056	0.174
lnsize	2140	22.830	1.540	19.970	21.720	22.620	23.810	27.040
TPGC	2140	0.749	0.156	0.465	0.597	0.757	0.867	1.000
RED	2140	5421.000	2245.100	1788.000	2912.000	4934.000	7511.000	9121.000
CP	2140	613.000	170.1	407.000	519.000	589.000	644.000	1037.000
GHI	2140	8.978	19.657	−29.340	0.000	0.000	5.000	133.600
UMC	2140	0.789	0.289	0.001	0.698	0.794	0.869	10.810

. It indicates that the minimum value of the CEI for enterprises is 0.059, the maximum value is 2.384. According to Table 1, there are significant differences in the intensity of carbon emissions among different types and regions. Since the implementation of the EPLPRC began in 2015, the sample data regarding the impact of the EEF on the CEI from the perspective of government intervention only includes 1885 observations.

3.4. Model Section

Environmental regulations, serving as policy tools to address market failures caused by the negative externalities of pollution, constitute binding exogenous constraints on power firms’ production and investment decisions. They also represent significant external factors influencing carbon emission levels. Since fixed-effects models are typically employed to analyze panel data observing individuals or units across multiple time points, we draw upon the work of Chen et al. (2020) [32] to construct the following model. The model aims to verify whether excess penalties and carbon pricing in environmental regulations influence corporate carbon emissions.

From the perspective of government intervention, this paper examines the impact of EEF on carbon emissions from power companies following the implementation of EPLPRC:

$${CEI}_{i,t}={\alpha }_0+{\alpha }_1{EEF}_{i,t}+{\alpha }_{2}Control{s}_{i,t}+{FirmEffect}_{i,t}+{TimeEffect}_{i,t}+{\epsilon }_{i,t}$$

(7)

From the perspective of market incentives, this paper examines the impact of CAP on carbon emissions from power companies following the launch of China’s ETS pilots:

$$\begin{gathered} {CEI}_{i,t}={\beta }_0+{\beta }_1{CAP}_{i,t}+{\beta }_2{CAP}_{i,t}^2+{\beta }_{3}Control{s}_{i,t}+{FirmEffect}_{i,t}+{TimeEffect}_{i,t} \\ +{\epsilon }_{i,t}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} \end{gathered}$$

(8)

Here, $i$ and $t$ denote firm and period; the dependent variable ${CE}_{i,t}$ represents the firm’s carbon emission intensity. ${TimeEffect}_{i,t}$ denotes the time fixed effect, ${FirmEffect}_{i,t}$ represents the firm fixed effect. ${\epsilon }_{i,t}$ signifies the random disturbance term. To determine whether the combined effects of China’s ETS pilots and EPLPRC are additive or mutually offsetting, we have established the following model.

$$\begin{array}{cc}\hfill {CEI}_{i,t}& ={\phi }_0+{\phi }_1{CAP}_{i,t}+{\phi }_2{CAP}_{i,t}^2+{\phi }_3{EEF}_{i,t}+{\phi }_4{CAP}_{i,t}\times pE{EF}_{i,t}\hfill \\ & +{\phi }_{5}Control{s}_{i,t}+{FirmEffect}_{i,t}+{TimeEffect}_{i,t}+{\epsilon }_{i,t}\hfill \end{array}$$

(9)

$$\begin{array}{cc}\hfill {CEI}_{i,t}& ={\phi }_0+{\phi }_1{CAP}_{i,t}+{\phi }_2{CAP}_{i,t}^2+{\phi }_3{EEF}_{i,t}+{\phi }_4{CAP}_{i,t}\times E{EF}_{i,t}+{\phi }_5{CAP}_{i,t}^2\times E{EF}_{i,t}\hfill \\ & +{\phi }_{6}Control{s}_{i,t}+{FirmEffect}_{i,t}+{TimeEffect}_{i,t}+{\epsilon }_{i,t}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hfill \end{array}$$

(10)

Theoretically, the synergistic effects generated by incentives and penalties can compensate for the limitations of individual regulations, exhibiting complementary advantages, cumulative incentives, and comprehensive enhancement. We introduced the interaction terms. Specifically, ${\phi }_4$ captures the linear interaction effects between the two approaches; ${\phi }_5$ captures whether the nonlinear effect of the CAP is influenced by EEF.

4. Empirical Results

4.1. Regression Results

Table 2. Empirical results.

	(1)	(2)	(3)	(4)
Variables	CEI
EEF	−1.444 ***		−1.162 **	−0.435 **
	(−4.69)		(−2.07)	(−2.46)
CAP		0.053 **	0.050 *	0.009 *
		(2.22)	(1.71)	(1.93)
CAP2		−0.001 ***	−0.001 ***	−0.000 ***
		(−2.77)	(−3.92)	(−3.26)
EEF×CAP			0.051 ***	0.021 **
			(3.14)	(2.26)
EEF×CAP2				−0.000 **
				(−2.48)
TPGC	−8.575	−7.015		2.363 ***
	(−0.89)	(−1.11)		(2.98)
RED	6.808	6.316		−2.112 ***
	(1.00)	(1.41)		(−3.14)
CP	0.002 **	0.001 *		−0.000 ***
	(2.17)	(1.75)		(−3.84)
ROE	8.532	2.869	−0.709	−0.248
	(1.07)	(0.56)	(−0.12)	(−0.28)
Lev	−7.679 ***	−2.572 ***	−1.905 ***	−0.089
	(−8.02)	(−4.26)	(−2.60)	(−0.47)
TobinQ	0.577 ***	0.108	−0.198 *	−0.061 **
	(3.72)	(1.06)	(−1.74)	(−2.24)
EPS	0.148	0.680 **	−0.816 **	−0.065
	(0.28)	(2.07)	(−2.06)	(−0.65)
ROA	63.436	34.080	43.439	1.411
	(0.62)	(0.53)	(0.57)	(0.19)
lnsize	3.069 ***	1.126 ***	−0.196	0.157 ***
	(25.52)	(13.34)	(−1.63)	(4.99)
Constant	−76.722 ***	−29.524 ***	5.270 *	−5.048 ***
	(−6.78)	(−4.02)	(1.95)	(−4.02)
Observations	1885	2140	1885	1885
R-squared	0.329	0.705	0.634	0.198
Firm FE	YES	YES	YES	YES
Time FE	YES	YES	YES	YES

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. Figures in parentheses present t-values.

Column (1) shows that from the perspective of government intervention, controlling for other factors, EEF exhibits a significant negative correlation with carbon emission intensity. This aligns with the widely held assumption that stricter penalties for excess emissions prompt firms to reduce emissions, reflecting the constraint effect of environmental regulations.

Column (2) of Table 2 reveals that from a market incentive perspective, CAP guides firms to reduce emissions through market mechanisms and price signals, but this carbon reduction effect is conditional. The impact of CAP on CEI follows an inverted U-shaped curve. At lower carbon prices, when carbon costs fall below firms’ marginal profits, firms tend to directly purchase required allowances, leading to increased rather than decreased carbon intensity. Once carbon prices exceeded the marginal abatement cost inflection point, firms shift toward voluntary emission reductions, energy efficiency improvements, fuel substitution, and investments in green technology innovation. Carbon emissions then decline as prices rise, resulting in an inverted U-shaped relationship where emissions first increase before decreasing.

Table 2 columns (3) and (4) clearly illustrate the interactive effects of EEF and CAP on corporate carbon emissions. The significance levels and sign directions of $E{EF}_{i,t}$, ${CAP}_{i,t}$, and ${CAP}_{i,t}^2$ align with univariate regressions, confirming the robustness of their main effects. Both policies significantly influence emissions: EEF directly suppresses emissions, while CAP exhibits an inverted U-shaped effect.

When examining CAP and EEF simultaneously, key changes in interaction terms reveal synergistic effects among these regulatory tools. In Column (3) of Table 2, the coefficient for ${CAP}_{i,t}\times E{EF}_{i,t}$ is significantly positive, while that for ${CAP}_{i,t}^2\times E{EF}_{i,t}$ is significantly negative, indicating a significant synergistic moderation effect between EPLPRC and China’s ETS pilots. On one hand, when CAP is low, EEF enhances the promotion effect on corporate CEI, increasing the magnitude of CEI growth. On the other hand, when CAP is high and exceeds the inflection point, EEF accelerates the emission reduction effect after the carbon price breaks through the inflection point, hastening the decline in CEI. Ultimately, EPLPRC steepens the inverted U-shaped curve between CAP and CEI and shifts the inflection point to the left. These findings indicate that the policy combination of EPLPRC and China’s ETS pilots not only enhances the regulatory efficiency of CAP but also lowers the carbon price threshold for emission reduction effects. This provides empirical support for synergistic regulation under dual environmental frameworks—government intervention and market incentives.

These findings offer critical insights for environmental regulation design: while maintaining the binding force of administrative penalties, flexibly leverage incentive mechanisms from market-based tools like carbon markets [32]. The balance between these two approaches should be dynamically adjusted according to the carbon market’s developmental stage to ensure maximized emission reduction outcomes.

4.2. Endogeneity Test

To mitigate potential endogeneity issues, we employ two approaches: two-stage least squares (2SLS) under the instrumental variable’s framework and within-group difference-in-differences.

From a market incentive perspective, we select the lagged international oil price (IV1) and lagged natural gas price (IV2) as instrumental variables for the impact of carbon pricing on power companies’ carbon emissions. On one hand, carbon prices exhibit significant interdependence with international oil prices and fossil fuel prices, influencing carbon pricing through energy substitution and marginal abatement costs—a linkage that intensifies under extreme risk conditions. Here, using lagged terms for IV1 and IV2 employs one-period-back energy prices to mitigate potential simultaneous reverse causality. This approach aligns with the exclusion requirement that instrumental variables influence firm emissions solely through carbon prices, as current carbon price changes cannot immediately alter previous-period oil or natural gas prices, and energy prices affect carbon price via cost-driven mechanisms influencing carbon allowance demand. Conversely, power companies’ carbon emissions are not directly influenced by the previous period’s international oil or natural gas prices.

The regression results are presented in

Table 3. Endogeneity test.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	First Stage		Second Stage	First Stage	Second Stage
	CAP	CAP2	CEI	EEF	CEI	CEI	CEI
CAP			0.277 **
			(2.000)
CAP2			−0.005 **
			(−2.420)
IV1	1.601 ***	105.570 ***
	(4.050)	(3.560)
IV12	−0.179 ***	−12.480 ***
	(−8.370)	(−7.610)
IV2	4.911 ***	460.287 ***
	(6.75)	(7.400)
IV22	−0.196 ***	−22.144 ***
	(−3.640)	(−5.030)
EEF					−2.603 ***
					(−3.0519)
IV3				−0.151 ***
				(−10.870)
diff_CAP						0.000 ***
						(19.676)
diff_CAP2						−1.093 ***
						(−3.695)
diff_EEF							−0.5009 ***
							(−5.573)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Stkcd FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	2117	2117	2117	1885	1885	2117	1885
Kleibergen–Paap rk Wald F	17.061 [7.56]			118.107 [16.38]		-	-
Kleibergen–Paap rk LM statistic	86.592 ***			94.578 ***		-	-

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. Figures in parentheses present t-values.

, columns (1), (2), and (3). The first-stage results indicate that the coefficients for IV1 and IV2 are significantly positive at the 1% level, while the coefficients for IV1² and IV2² are significantly negative at the 1% level. The Kleibergen–Paap rk LM test yields a p-value of 0.000, and the F-statistic exceeds 10%, confirming that neither instrument IV1 nor IV2 suffers from weak instrumentality issues. The second-stage results indicate that the estimated coefficients for CAP and CAP² are significant at the 5% level, with magnitudes and signs consistent with the benchmark regression results. This confirms that after addressing partial endogeneity using instrumental variables IV1 and IV2, carbon prices still significantly influence power companies’ carbon emission intensity, fully aligning with the preceding conclusions.

From a government intervention perspective, the number of enforcement and penalty cases under EPLPRC (IV3) across provinces was selected as an instrumental variable for the law’s impact on power companies’ carbon emissions. On one hand, the number of enforcement and penalty cases in each region is highly correlated with corporate environmental compliance and emission reduction behaviors, satisfying the requirement for relevance. Furthermore, power companies’ carbon emissions are not directly affected by IV3.

The regression results are presented in Table 3, columns (4) and (5). The first-stage results indicate that the coefficient of IV3 is significantly negative at the 1% level. The Kleibergen–Paap rk LM test yields a p-value of 0.000, and the F-statistic exceeds 10%, confirming that the instrumental variable IV3 does not suffer from weak instrumentality issues. The second-stage results show that the estimated coefficient for EEF is significantly negative at the 1% level, consistent with the benchmark regression results. This indicates that after addressing partial endogeneity using IV3, EPLPRC still significantly suppresses carbon emission intensity in power enterprises, aligning with the earlier conclusion.

Furthermore, we employed an intra-group difference transformation to address endogeneity issues. The regression results are shown in columns (6) and (7) of Table 3. The coefficient for diff_EEF is negative, the coefficient for diff_CAP is significantly positive, and the coefficient for diff_CAP² is significantly negative, all at the 1% level. This indicates that after addressing endogeneity using the intra-group difference method, H1 and H2 remain valid. The endogeneity test is passed.

4.3. Robustness Test

4.3.1. Eliminating the Interference of Other Competitive Policies

Other policies implemented during the same period may also influence the carbon emission intensity of thermal power enterprises, such as electricity market reforms, sharp fluctuations in fossil fuel prices, and the Air Pollution Prevention and Control Action Plan (APPCA). The implementation of these policies could confound our research findings, making it impossible to determine the true impact of carbon pricing on thermal power enterprises’ carbon emissions. Therefore, to eliminate interference from other relevant policies during the same period, we employ policy dummy variables did_ele, did_coalprice, and did_APPCA to control for the exogenous shocks of these policy events. After incorporating these policy dummies into the baseline model, the regression results are presented in

Table 4. Robustness tests.

PANEL A	(1)	(2)	(3)	(4)	(5)	(6)
	CEI_did_ele		CEI_did_coalprice		CEI_did_APPCA
EEF	−0.201 ***		−0.216 ***		−0.231 ***
	(−5.62)		(−6.11)		(−6.83)
EEF×did	0.047		0.037		−0.618 ***
	(0.56)		(0.39)		(−2.78)
CAP		0.116 ***		0.122 ***		0.120 ***
		(3.72)		(3.63)		(4.31)
CAP×did		−0.086 *		−0.084		−0.118
		(−1.69)		(−1.45)		(−1.61)
CAP2		−0.002 ***		−0.002 ***		−0.002 ***
		(−4.95)		(−4.72)		(−5.42)
CAP2×did		0.001 **		0.001 *		0.002 **
		(2.14)		(1.92)		(2.09)
Constant	0.468 ***	−1.394 ***	0.490 ***	−1.560 ***	0.524 ***	−1.596 ***
	(20.24)	(−2.65)	(21.42)	(−2.69)	(23.48)	(−3.57)
Controls	YES	YES	YES	YES	YES	YES
Obs.	1855	2140	1855	2140	1855	2140
R-squared	0.170	0.626	0.156	0.627	0.141	0.627
Firm FE	YES	YES	YES	YES	YES	YES
Time FE	YES	YES	YES	YES	YES	YES
PANEL B	(7)	(8)	(9)	(10)	(11)	(12)
	CEI
EEF	−3.763 ***				−2.341 ***
	(−2.67)				(−4.76)
CAP		0.013 **				0.142 **
		(2.03)				(2.39)
CAP2		−0.000 ***				−0.004 ***
		(−2.67)				(−5.08)
L.EEF			−0.400 ***
			(−3.03)
L.CAP				0.121 ***
				(2.94)
L.CAP2				−0.003***
				(−5.36)
Constant	−0.606 ***	−3.341 ***	−16.058	2.491 ***	−84.556 ***	−59.340 ***
	(−18.51)	(−6.75)	(−1.38)	(3.75)	(−4.41)	(−16.26)
Controls	YES	YES	YES	YES	YES	YES
Obs.	781	1051	1631	1929	780	928
R-squared	0.049	0.093	0.147	0.052	0.417	0.397
Firm FE	YES	YES	YES	YES	YES	YES
Time FE	YES	YES	YES	YES	YES	YES

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. Figures in parentheses present t-values.

, Panel A, columns (1)–(6). In summary, the benchmark regression results remain robust after accounting for other policy interventions. This confirms that carbon pricing indeed exerts an inverted U-shaped effect on the carbon emission intensity of thermal power plants, rather than being driven by other policies, thereby passing the robustness test.

4.3.2. Changing the Sample Period

The regression results presented earlier are based on the full sample data from 2013 to 2021. During the sample period, the occurrence of major random events may have reduced power generation, thereby lowering the carbon emission intensity of thermal power enterprises. Therefore, to avoid the impact of major unforeseen events during the sample period, data from 2021–2024—after the formal launch of China’s carbon market—were selected for verification. The results are presented in Table 4, PANEL B, columns (7) and (8). The regression results indicate that when the sample period is shifted to the operational phase of China’s carbon market, an inverted U-shaped relationship between carbon prices and power companies persists, consistent with the conclusions from the previous regression analysis.

4.3.3. Adding Lagged Variables

To eliminate the possibility of the dependent variable CEI exerting a reverse effect on the explanatory variables, we applied lagged treatment to the core explanatory variables to address potential reverse causality issues. As shown in columns (9) and (10) of Table 4 PANEL B, the positive or negative nature and significance of the core explanatory variables are consistent with the results of the benchmark regression Equations (6) and (7). Robustness tests were passed.

4.3.4. Reducing Study Sample

To avoid the interference of energy type heterogeneity among power enterprises on the benchmark regression results, we retained thermal power enterprises participating in both the pilot carbon market and the national carbon market for regression analysis. Table 4 PANEL B shows that the coefficient for explanatory variable EEF in column (11) is −2.341, while the coefficients for ${CAP}_{i,t}$, and ${CAP}_{i,t}^2$ in column (12) are 0.142 and −0.004, respectively, both statistically significant. This indicates that the benchmark conclusions remain robust.

4.4. Mechanism Analysis

4.4.1. The Intermediary Role Test of Unit Material Cost

The EU Emissions Trading Scheme reflects the costs of complying with emission caps or parallel renewable energy incentive programs through its impact on unit material costs, thereby significantly affecting the power and cement industries [10]. EPLPRC increases coal consumption per unit output through factor substitution or innovation compensation effects, consequently lowering the CEI [9].

The core source of carbon emissions for power enterprises is fossil fuel consumption, with fuels like coal serving as core production materials. In the unit material costs of thermal power plants, coal for power generation accounts for an extremely high proportion, and its consumption intensity directly determines total carbon emissions. Simultaneously, carbon emissions from raw material extraction, transportation, and other stages are also transmitted through material costs into corporate emissions reduction decisions. Changes in unit material costs reflect both the impact of carbon pricing on production inputs and directly influence the CEI. In the power sector, carbon pricing makes carbon costs explicit through quota constraints and trading. Given the high proportion of fuel costs in coal-fired power plants, carbon constraints alter unit dispatch and fuel input structures, thereby affecting output per unit.

The benchmark regression only demonstrates that government intervention and market incentives have a significant impact on carbon emissions from power companies. Further examination is needed to identify the specific channels through which these factors influence carbon emissions. To achieve effective causal inference, we employ the two-stage model stepwise regression analysis to examine the channeling effect of UMC, constructing the following equation [33].

Phase One:

$${UMC}_{i,t}={\alpha }_0+{\alpha }_1{EEF}_{i,t}+{\alpha }_{2}Control{s}_{i,t}+{FirmEffect}_{i,t}+{TimeEffect}_{i,t}+{\epsilon }_{i,t}$$

(11)

Phase Two:

Use ${CEI}_{i,t}$ regression to fit ${EEF}_{i,t}$ and ${\widehat{UMC}}_{i,t}$ obtained in the first stage.

$${CEI}_{i,t}={\alpha }_0+{\alpha }_1{\widehat{UMC}}_{i,t}+{\alpha }_{2}Control{s}_{i,t}+{FirmEffect}_{i,t}+{TimeEffect}_{i,t}+{\epsilon }_{i,t}$$

(12)

Among these, coefficient α₁ estimates the impact of the EEF on the CEI of power enterprises through the UMC. As shown in columns (1) and (2) of

Table 5. Mechanism analysis.

	(1)	(2)	(3)	(4)
Variables	UMC	CEI	GTI	CEI
UMC_hat	0.027 ***
	(3.18)
CAP		−52.355 ***
		(−4.54)
GTI_hat			−0.039 **
			(−2.26)
GTI_hat2			0.001 **
			(2.52)
CAP				2.827 **
				(2.23)
CAP2				−0.458 ***
				(−2.87)
Controls	YES	YES	YES	YES
Constant	0.843 ***	−24.278 *	28.899 ***	55.930 **
	(2.90)	(−1.79)	(5.94)	(2.36)
Observations	1907	1885	1795	2140
R-squared	0.161	0.327	0.180	0.344
Firm FE	YES	YES	YES	YES
Time FE	YES	YES	YES	YES

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. Figures in parentheses present t-values.

, the EEF effectively curbed carbon emissions through the UMC channel.

4.4.2. The Intermediary Role Test of Green Technology Innovation

The Porter hypothesis (Porter and van der Linde, 1995) [16] posits that well-designed environmental regulations can exert external pressures that drive firms to engage in technological innovation activities. Research by Ali et al., (2024) and Zhang et al., (2024) indicates that CAP significantly enhances firms’ green technological innovation capabilities [34,35]. They found that market-based environmental regulations foster green innovation by accelerating technology diffusion, equipment renewal and upgrades. However, a crucial prerequisite for the innovation compensation effect of market-incentive environmental regulations to take effect is appropriately calibrated environmental regulation. Both excessively weak and excessively stringent environmental regulations may hinder corporate technological innovation. Weak regulations reduce the environmental responsibilities and costs borne by enterprises, minimizing or even eliminating their impact on business operations, thereby diminishing firms’ motivation for proactive innovation. Conversely, overly stringent environmental controls may force enterprises to exit the market if they fail to meet environmental standards.

Green technological innovation is pivotal to achieving green economic development [36,37]. The impact of green technological innovation on carbon emissions is unidirectional: once innovations are implemented, they continuously optimize energy consumption structures, eliminating the risk of reverse causality where carbon emission changes determine green technological innovation. Enterprises that fail to reduce carbon emissions face severe consequences such as heavy fines and production suspensions for rectification. Green technological innovation serves as the core pathway for achieving carbon reduction. Specifically, power companies reduce carbon emissions through green innovation channels including R&D in renewable energy generation technologies, ultra-low emission coal-fired power technologies, ultra-supercritical power generation technologies, and coal-fired power combined with carbon capture, utilization, and storage (CCUS) technologies.

To achieve effective causal inference, we employ the two-stage model to examine the channeling effect of Green Technology Innovation (GTI), constructing the following equation.

Phase One:

$${GTI}_{i,t}={\beta }_0+{\beta }_1{CAP}_{i,t}+{\beta }_2{CAP}_{i,t}^2+{\beta }_{3}Control{s}_{i,t}+{FirmEffect}_{i,t}+{TimeEffect}_{i,t}+{\epsilon }_{i,t}$$

(13)

Phase Two:

Use ${CEI}_{i,t}$ regression to fit ${CAP}_{i,t} , {CAP}_{i,t}^2$ and ${\widehat{GTI}}_{i,t}$ obtained in the first stage.

$${CEI}_{i,t}={\alpha }_0+{\alpha }_1{\widehat{GTI}}_{i,t}+{\alpha }_2{\widehat{GTI}}_{i,t}^2+{\alpha }_{3}Control{s}_{i,t}+{FirmEffect}_{i,t}+{TimeEffect}_{i,t}+{\epsilon }_{i,t}$$

(14)

As shown in columns (3) and (4) of Table 5, among the inverted U-shaped effects of CAP on power enterprises, GTI is an effective channel of influence.

Changes in low-carbon technology innovation investment exhibit a U-shaped pattern with distinct phases, serving as a crucial link between carbon pricing and emissions. When carbon prices are low, enterprises face minimal pressure to reduce emissions. At this stage, the returns on low-carbon technology innovation fall below the costs, leading companies to prefer maintaining high-emission production models. Innovation investment remains low, exerting virtually no restraint on emissions, which grow in tandem with production demand. As carbon prices gradually rise toward the critical threshold, enterprises recognize the need for emissions reduction but still prefer purchasing allowances over investing in innovation. Technological innovation investment grows slowly, and carbon emissions continue to trend upward. Once carbon prices surpass the critical threshold, the cost of purchasing allowances far exceed R&D investment. Companies then significantly increase low-carbon technology innovation spending—such as upgrading desulfurization and denitrification technologies in coal-fired power plants or developing new energy generation technologies. Once implemented, these technologies rapidly curb carbon emissions, propelling them into a downward trajectory.

4.5. Heterogeneity Analysis

4.5.1. Heterogeneity in Firm Size and Property Right

Considering differences in property rights characteristics, firm size, and regional resource endowments, this study examines the heterogeneous effects of EEF and CAP on the CEI within an environmental regulatory framework.

First, we examine heterogeneity in property rights and firm size. From a government intervention perspective,

Table 6. Heterogeneity analyses.

Panel A	Large-SOEs	Small-SOEs	Large-NSOEs	Small-NSOEs
	(1)	(2)	(3)	(4)
EEF	−1.398 **	0.009	0.320	−0.105 ***
	(−2.55)	(0.38)	(1.31)	(−2.70)
Constant	25.988	−2.289 ***	−9.751	0.602
	(1.41)	(−2.81)	(−0.96)	(0.65)
Controls	YES	YES	YES	YES
Obs.	553	518	178	636
R-squared	0.587	0.105	0.309	0.204
Firm FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Panel B	Large-SOEs	Small-SOEs	Large-NSOEs	Small-NSOEs
	(5)	(6)	(7)	(8)
CAP	0.059 *	0.006	−0.158 **	−0.001
	(1.77)	(1.46)	(−2.48)	(−0.36)
CAP2	−0.001	−0.000	0.004 ***	0.000
	(−1.45)	(−0.96)	(3.18)	(0.66)
Constant	12.317	0.609	−37.912 *	1.507
	(0.42)	(0.51)	(−1.82)	(1.49)
Controls	YES	YES	YES	YES
Obs.	634	571	216	719
R-squared	0.330	0.327	0.735	0.176
Firm FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. Figures in parentheses present t-values.

PANEL A reveals that EEF significantly suppresses CEI for both small non-state-owned power enterprises and large state-owned power enterprises. This stems from state-owned enterprises (SOEs), as core policy implementers, facing stronger political accountability and environmental compliance pressures. The EEF not only directly impacts corporate financial performance but is also closely tied to the performance evaluations of SOE executives. Small non-state-owned enterprises, characterized by smaller scale and lower profit margins, bear a higher proportion of EEF costs relative to their operating expenditures compared to large SOEs. This results in stronger constraints on production decisions due to compliance costs. For enterprises choosing not to reduce emissions, the EEF may directly squeeze profit margins, making its suppressing effect equally significant for small non-state-owned enterprises.

4.5.2. Heterogeneity in Thermal Power Installed Capacity

The heterogeneity in the growth rate of thermal power generation capacity fully reflects regional disparities in electricity supply demand pressures and energy structure transformation. According to factor substitution theory, when environmental regulations tighten, enterprises often seek alternative factors to reduce reliance on high-pollution, high-emission inputs. If a region experiences rapid growth in thermal power generation capacity, it indicates that thermal power will remain dominant in the short term.

Focusing solely on the national average growth rate of thermal power capacity risks obscuring regional heterogeneity. Some provinces accelerate thermal power construction due to load growth, insufficient external power supply, or local heating demands; others slow down because of favorable renewable energy integration conditions and ample cross-regional transmission capacity. Therefore, we examine the heterogeneity in power generation capacity growth across different regions to avoid misinterpreting local signals through national averages.

We categorize the sample into low-growth and high-growth groups based on regional thermal power capacity growth rates. As shown in

Table 7. Heterogeneity analyses.

	(1)	(2)	(3)	(4)
Variables	Slowgrowth	Rapidgrowth	Slowgrowth	Rapidgrowth
CAP			−0.021	0.115 ***
			(−0.76)	(2.97)
CAP2			0.000	−0.002 ***
			(0.97)	(−4.45)
EEF	−0.569 ***	0.131
	(−3.49)	(1.13)
Constant	−4.317 *	−2.206	0.216	−1.614 **
	(−1.78)	(−0.84)	(0.55)	(−2.25)
Controls	YES	YES	YES	YES
Obs.	890	995	1029	1111
R-squared	0.284	0.253	0.456	0.647
Firm FE	YES	YES	YES	YES
Time FE	YES	YES	YES	YES

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. Figures in parentheses present t-values.

, the EEF exhibits a more pronounced suppression effect on the CEI in regions with slower thermal power capacity growth. These regions primarily rely on existing thermal power plants, where renewable energy substitution is well-established. The penalty pressure from the EEF can be rapidly transmitted through environmental upgrades of existing units and renewable energy substitution and local policy enforcement is stricter. Such regions have typically passed their large-scale expansion phase, with relatively stable unit structures and personnel systems. Consequently, the policy transmission chain for the EEF is shorter, implementation costs are lower, and policy effects are more direct.

In regions with rapid thermal power capacity growth, CAP and CEI exhibit a more pronounced inverted U-shaped relationship. When carbon prices are low, enterprises expand production by adding new thermal power units to pass on carbon costs, leading to increased emission intensity. Once carbon prices surpass the inflection point, rigid carbon cost constraints force new units to undergo environmental upgrades and integrate renewable energy, enabling rapid emission intensity reduction through the cost advantages of scaled-up emissions reduction. When thermal power capacity growth slows and supply-demand tensions ease, operational pressures shift toward efficiency and cost competition. At this stage, refined improvements in energy efficiency and management become the primary focus, and the incentives and constraints of EEF more readily translate into short-term reductions in CEI.

These findings indicate that the emission reduction effects of EEF and CAP are highly dependent on regional energy development stages. The conclusions above provide empirical support for formulating differentiated environmental regulatory policies by region.

4.5.3. Further Analysis

Further analysis of the dual policy regulation reveals that for power companies in regions with slower growth in thermal power capacity, the synergistic effects between EEF and CAP are more pronounced. In this context, the significance and signs of ${CAP}_{i,t}$, ${CAP}_{i,t}^2$, and $E{EF}_{i,t}$ remain unchanged, while the interaction term ${CAP}_{i,t}\times E{EF}_{i,t}$ is significantly positive and ${CAP}_{i,t}^2\times E{EF}_{i,t}$ is significantly negative.

Table 8. Further analysis.

Variables	(1)	(2)
	Slowgrowth	Rapidgrowth
avgprice	0.137 **	0.083
	(2.28)	(0.99)
avgprice2	−0.002 **	0.002 *
	(−2.43)	(1.74)
CCER81201	−5.385 ***	−6.349 *
	(−2.88)	(−1.88)
CCER81201avgprice	0.275 **	0.293
	(2.41)	(1.63)
CCER81201avgprice2	−0.004 **	−0.001
	(−2.40)	(−0.37)
Constant	−0.998	−1.802
	(−1.11)	(−1.06)
Controls	YES	YES
Obs.	890	995
R-squared	0.025	0.105
Firm FE	YES	YES
Time FE	YES	YES

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. Figures in parentheses present t-values.

results indicate that for power companies in regions with slower growth in thermal power capacity, EEF and CAP exhibit more pronounced synergistic effects. When CAP is low, EEF curbs firms’ impulse to pass on carbon costs through production expansion by imposing dual constraints of carbon costs and penalty costs, thereby flattening the upward slope of the inverted U-shaped curve. After CAP crosses the inflection point into the high-price zone, EEF’s penalty pressure increases the marginal benefits of emissions reduction for enterprises. Combined with the cost advantages of sufficient renewable energy substitution, this rapidly translates emissions reduction incentives of CAP into actual actions, resulting in a steeper downward slope of the inverted U-shape.

At this stage, the synergistic effect of the dual policies amplifies emissions reduction elasticity of CAP, enhancing policy regulatory efficiency. This finding indicates that the synergistic effects of dual policy regulation are highly dependent on a region’s stage of energy development. Against the backdrop of stock optimization during transition, the coordination between EEF and CAP effectively compensates for the shortcomings of single policies, providing crucial empirical evidence for designing and implementing differentiated policy combinations at the regional level.

5. Conclusions

Among various factors, the excessive consumption of traditional fossil fuels by thermal power enterprises is the primary driver behind persistently high carbon emissions. Advancing the clean energy transition has become a critical pathway to achieving sustainable development and emissions reduction—for both developed and developing countries alike. This underscores the pivotal role of environmental regulations in guiding energy structure transformation and reducing carbon emissions.

This study adopts a dual perspective of government intervention and market incentives. Using theoretical models and empirical analysis based on panel data from China’s power sector (2013–2021), it systematically evaluates the synergistic effects of the EPLPRC and Pilot ETS on the CEI of power enterprises. It further examines heterogeneity and transmission channels, yielding several key conclusions: First, it demonstrates the significant suppression effect of the EEF on power companies’ carbon emissions and the inverted U-shaped relationship between CAP and carbon emissions. Based on this, it discusses the synergistic effects between the two. Second, unit coal consumption cost and green technological innovation emerge as key channels through which EPLPRC and Pilot ETS generate carbon reduction effects. Third, it further reveals that under different thermal power installation growth scenarios, energy structure transformation effectively modulates the carbon reduction effects of EPLPRC and Pilot ETS.

This paper proposes the following policy recommendations: First, empirical evidence indicates that EPLPRC contributes to carbon reduction in power enterprises. Therefore, while responding to global temperature limit goals and national carbon neutrality targets, local governments should strengthen environmental enforcement, properly exercise administrative penalty authority, compel polluting enterprises to undertake green transformation, and promote the development of high-pollution enterprises. The central government should actively supervise local environmental enforcement and conduct inspection campaigns to ensure effective implementation of the National Environmental Protection Law. Second, green innovation is an indispensable channel for carbon markets to reduce the CEI. Beyond penalizing polluters, local governments should offer tax rebates and incentives to enterprises demonstrating outstanding green transformation, while providing subsidies and rewards to polluting enterprises actively investing in green innovation. Third, coal consumption per unit output serves as an effective channel for the EPLPRC to curb CEI. Governments should bolster support for infrastructure like clean energy generation capacity to encourage thermal power enterprises to proactively transition to clean energy. Finally, local governments should prioritize addressing heterogeneity in environmental regulation enforcement intensity to fully leverage the synergistic effects of dual environmental regulations.

We identify current research limitations from two dimensions: sample data characteristics and methodological design. First, sample data limitations stem from selective bias in coverage. Our focus on listed power companies at the micro level overlooks significant differences in emission reduction elasticity compared to unlisted small-to-medium enterprises. Listed firms, particularly large state-owned enterprises, typically possess stronger policy responsiveness and foundational capabilities for green technological innovation, potentially leading to underestimation of environmental regulations and carbon pricing impacts on this group. Second, the precision of coal consumption cost per unit output data is limited. Existing data primarily relies on energy cost indicators from annual corporate financial statements, lacking breakdowns by coal procurement costs, coal transportation costs, and coal loss costs. This hinders the accuracy of mechanism identification. Methodologically, spatial spillover effects were not fully considered—carbon price fluctuations in pilot regions may influence power dispatch and carbon emission behaviors of enterprises in neighboring regions through inter-provincial power transmission.

In subsequent research, we will strive to expand the sample size, refine the measurement standards for mechanism variables, and supplement micro-level behavioral data to provide more direct micro-level evidence for further mechanism analysis. Methodologically, we will consider constructing spatial econometric models to address spillover effects. Using “inter-provincial electricity transmission volume” as the weight matrix, we will quantify the differences in inflection points of the inverted U-shaped relationship between carbon prices and carbon emissions under various scenarios, thereby enhancing the depth of heterogeneity analysis.