Identifying the Impact of Climate Policy on Urban Carbon Emissions: New Insights from China’s Environmental Protection Tax Reform

Abstract

Environmental protection tax (EPT), as a major tool to improve air quality and reduce carbon emissions, is of great significance for promoting urban low-carbon transformation. In this context, this paper has compiled a dataset from 282 Chinese cities during 2006–2022 and empirically identify the implication of EPT for carbon emissions at the city level by using the intensity difference-in-differences (I-DID) model. The result discloses that EPT greatly lowers carbon emissions by an average of 10.9% compared to non-pilot cities. Even after conducting some robustness checks, the result remains unchanged. Mechanism testing reveals that EPT curbs carbon emissions through enhancing energy utilization efficiency, fostering green technological advancements, and modernizing urban industries. Meanwhile, we show that EPT exerts a more substantial effect on carbon emissions in innovative cities, central and western cities, non-industrial-based cities, and non-resource-dependent cities. More importantly, EPT greatly promotes imitation and learning in neighboring regions, forming a radiation impact upon carbon reduction in surrounding areas. Hence, these results offer an important decision-making guide for optimizing the EPT system, strengthening the coordinated governance of carbon emission across regions, and ultimately promoting urban low-carbon development.

1. Introduction

Climate change exhibits a serious and long-term dilemma facing human society. Notably, the sustained rise in greenhouse gas emissions mainly composed of carbon dioxide is having widespread and profound impairments for socioeconomic actions and human well-being [1]. According to 2024 Global Climate Status released by World Meteorological Organization (WMO), the global average temperature in 2024 is 1.55 °C higher than that during 1850–1900, making it the hottest year on record. Global warming, for example, leads to glacier melting and ocean expansion, causing a series of weather events such as heat waves, droughts, and floods [2]. Faced with severe climate situation, seeking plans to lessen carbon emissions has emerged as a critical concern around the world, especially for big developing countries [3]. Since 1978, the extensive growth mode for Chinese economy has triggered a sharp usage of fossil fuels and huge carbon emissions, and the paradox between economic growth and environmental control, all the more so, remains increasingly acute. According to 2025 World Energy Statistical Yearbook, as the global largest carbon emitter, China’s total carbon emissions reached 11.1728 billion tons in 2024, accounting for 31.5% of the world. Faced with the challenge of global climate change, China is striving to fulfill its international obligations and actively taking measures to curb carbon emissions [4]. In 2020, China pledged to attain carbon peak by 2030 and carbon neutrality by 2060 (hereinafter termed as the dual carbon target), demonstrating China’s determination and perseverance in addressing global warming. In 2022, the 20th Chinese National Congress, all the more so, proposed to follow the green and low-carbon development, in line with China’s energy and resource endowment, and realize the carbon peak action in a planned manner. In this context, paying more attention to China’s carbon emissions is crucial for addressing global climate change, formulating relevant measures, and promoting sustainable development [5]. Thus, how to greatly lessen carbon emissions at a regional level while attaining economic growth, improving ecological environment, as well as advancing sustainable development, has proved to be a key issue urgently needing to be addressed by governments and academia [6].

In light of global climate regulation and China’s dual carbon initiative, the nation is aggressively cutting carbon by deploying a variety of policies. These range from crafting a national carbon trading scheme to enhancing the green finance landscape and executing eco-friendly carbon sequestration programs. As a key tool of the green tax system, the environmental protection tax (EPT) notably exhibits a powerful force for attaining low-carbon development [7]. To be specific, China’s Environmental Protection Law, debuted in 2018, which emerged as the first eco-centric single-tiered tax legislation, represents a significant leap in tax system reforms [8]. Theoretically, the EPT law affects carbon emissions through a dual track involving price signals and incentivizing policies. Conversely, imposing taxes on pollutants raises the price tag on fossil fuels, making them less appealing on the market, and realizes a shift towards renewable energy [9]. Moreover, due to a nuanced tax form, an incentivized penalty system arises, hitting hard on polluting firms with higher emission taxes, while offering subsidies and tax exemptions for investing in green innovation [10,11]. This practice pushes firms into streamlining their production techniques, ultimately reducing local carbon emissions. Meanwhile, the combined impact of these tactics has the potential to dramatically curb local emissions level. Therefore, this study delves into these pivotal inquiries. For example, how does EPT execution affect Chinese carbon emissions at an urban scale? Is this impact realized via channels like industrial upgrading or technology innovation? Do the varying traits across cities lead to disparities in the effectiveness of EPT? Addressing these questions will not only augment the existing theories surrounding environmental control and low-carbon progress, but will also offer practical insights for optimizing emissions reduction strategies [12].

The research focuses on how the EPT scheme affects Chinese carbon emissions at the city level. To be precise, by adopting the intensity difference-in-differences (I-DID) technique, we collect a dataset from 282 Chinese cities during 2006–2022 and identify the implication of EPT for urban carbon emissions. In relation to earlier studies, the innovations of this study could be summed into four aspects. Firstly, by gathering the underlying data from five energy sources, we create a comprehensive carbon accounting framework at the city level, which helps to enhance the verity for assessing carbon emission. It not only clarifies the trend in Chinese carbon emissions at a city scale but also offers a guide to some extent for Chinese authorities to optimize carbon reduction strategies. Secondly, this study cleverly applies the I-DID model, which helps to deeply study the dynamic changes in policy execution and scientifically assess the impact of environmental regulations, to disclose the effects of EPT in reducing carbon emissions. The technique effectively compensates for the shortcomings of traditional static analysis and expands the scope of investigation into these driving factors for carbon emissions. Thirdly, we build a mathematical model, which helps to scientifically exhibit the economic behavior process of corporate carbon emissions through standardized analysis, to elucidate the role of EPT in curbing urban carbon emissions. Finally, we form the impact paths at a mechanism level, revealing the complex operation of EPT’s impact on carbon discharges, which facilitates better exertion of political emission reduction effects. By disclosing the impacts through three different channels, including energy efficiency, green technology, and industrial upgrading, we offer empirical evidence as well as underpin the theory behind effective ecological governance.

The rest of this study, in terms of structure, is formatted as follows. Section 2 delves into the prior literature relating to carbon emissions and EPT. Section 3 identifies how EPT affects carbon emissions at an urban scale and puts forth some hypotheses to be validated. More deliberately, Section 4 outlines the econometric model and dataset used for the empirical study. Section 5 presents the empirical outcomes coupled with their analysis. Lastly, we conclude the study and offer some policy plans.

2. Literature Review

This paper delves into the existing literature surrounding EPT and carbon emissions, offering a comprehensive summary of the present study status in this domain. The section primarily focuses on various factors affecting carbon emissions, economic and ecological effects of environmental regulation, and the use of DID technique across economics and ecology. What follows is an assessment of the research pertinent to the subject.

2.1. Factors Affecting Carbon Emissions

Longstanding interest within the academic realm has focused on disclosing the myriad elements that contribute to carbon emissions. After all, CO₂ is a key culprit in the ongoing climate crisis, and its presence endangers both our ecological systems and biodiversity. Moreover, it has far-reaching implications for human existence and advancement. Hence, scholars have poured over this issue extensively. Currently, researchers are scrutinizing multiple factors behind carbon discharges from several angles: economic expansion, industry make-up, technological progress, and global integration [2,13,14]. To delve deeper, it is widely recognized that growth in the economy often correlates with a spike in carbon emissions. Consequently, measures for curbing emissions are crucial for fostering a sustainable path forward. For illustration, Wang et al. [15] conducted an analysis using a fixed effects model on data spanning 2003–2019 across 30 Chinese provinces, revealing a clear link between economic development and carbon emissions. Alam and Hossain [16] investigated the association of economic expansion with CO₂ emissions by analyzing Chinese time-series data spanning 1990 to 2019. Employing unit root tests and dynamic OLS regression techniques, their research revealed a strong, positive long-term association between GDP-driven economic progress and rising carbon emissions. The findings demonstrated that as the economy grew, so did environmental pressures through heightened emissions. Nonetheless, a number of academics argue that over time, the effect of economic expansion on carbon emissions tends to form an inverted U-shape. For instance, the study by Mouthinho et al. [17], after examining the dataset from 13 sectors across Portugal and Spain from 1975 to 2012, concluded that economic growth and emissions display this inverted U-shape pattern. From an industrial standpoint, it is thought that shifting the industrial landscape by fostering new industries and nurturing high-tech companies can significantly cut down on the energy reliance of traditional sectors and diminish carbon emissions. For example, Xuan et al. [18] leveraged a dataset spanning 274 Chinese cities across 2011–2020 to demonstrate that the carbon-cutting potential of digital finance, under the backdrop of financial regulations, can be realized through driving industrial advancement. Yet, Shi et al. [5] present a contrasting viewpoint. Their analysis, utilizing data from 282 Chinese cities covering 2015–2020 and the Coupling Coordination Degree (CCD) technique, revealed that the outcome for industrial structure at the city level on carbon discharges is negligible. From a technological view, there is a lot of space when it comes to cutting down on carbon footprints. By leveraging digital technology, AI, and other forceful means, we can beef up energy efficiency, push for cleaner power solutions, and revamp industrial operations. To be precise, Gielen et al. [19] crunched a batch of renewable energy data and concluded that innovation is the driving force behind energy transition, and these technologies can greatly curb carbon emissions. Zhang et al. [20] dived into China’s manufacturing sector’s carbon footprint from 1990 to 2012, using the global meta-frontier directional distance function (GMN-DDF) method, and pinpointed that to really slash emissions, the government needs to champion innovation on the local scene and target resource-heavy sectors. When it comes to globalization, scholars are mainly split on how openness affects emissions. While some scholars assert technology transfers and stricter environmental norms can bring it down, others argue that more trade and the labyrinthine nature of global supply chains could actually hike it up. The study from Jijian et al. [21], which utilized data spanning from 1993 to 2018, revealed that a country’s imports tend to be positively associated with carbon emissions, whereas exports show a negative association with carbon emissions. As research advances and individual-level data becomes increasingly accessible, scholars have begun examining carbon emission drivers from a micro-level lens [22]. A notable study by Yang et al. [23] leveraged a sample from Chinese A-share industrial enterprises, spanning the years 2011–2019, to investigate how digital transformation impacts carbon footprints. Their findings reveal that corporate digitalization fosters low-carbon technological breakthroughs and eases carbon cycle strain through enhanced data analytics and operational efficacy, with particularly pronounced effects observed in low-emission enterprises.

2.2. The Implications for Environmental Regulations on the Economy and Environment

During more recent times, this rollout of a host of eco-friendly initiatives has made a major splash in both economy and environmental control. The academia has been all over this topic, diving deep into it with thorough research and spirited debate. On the one hand, the academic community holds four main perspectives on the economic effects of green governance [24,25,26]. For one thing, it is widely held that there is a strong relation between environmental regulations and economic growth [27]. Furthermore, a notable chunk of the scholarly crowd posits that ecological laws alongside economic performance exhibit an inverse U-shaped association, which initially hampers growth but ultimately fuels it. Chen et al. [9] discovered that there is this thing called an inverted U-curve when it comes to how environmental regulations affect economic growth, using dataset from a group of provinces in China over two decades. Conversely, Abdullah and Morley [8] conducted their study in the EU, analyzing data from 25 countries over a decade, and reported that raising environmental taxes does not lead to a big economic shake-up. Then there is this school of thought that thinks stricter eco-rules can actually fine-tune our industries with tech and more. Zhou et al. [28] employed spatial econometric modeling with a customized spatial weight matrix for the Yangtze River Delta region, demonstrating that enhanced environmental regulatory measures contribute to industrial upgrading. Furthermore, there is a general consensus that stricter environmental regulations can bolster the efficiency of eco-friendly production. In a study issued in 2022, Cheng and Kong [29] delved into a 2000–2019 dataset encompassing 30 provinces across China, employing a sophisticated spatial dynamic panel model. They discovered that a combination of policies tends to be more potent in spurring growth in green total factor productivity compared to a one-shot approach. Additionally, academics have embarked on highlighting the potential effects of environmental regulations on the job market. The research from Raff and Earnhart [30] suggests that enforcing environmental regulations, paradoxically, can lead to decreased production and environmental sector employment. Meanwhile, academic scrutiny of the governance impacts of such laws underscores their role in curbing and mitigating environmental harm and pollution [31,32]. For example, using 2005–2015 data from 30 Chinese provinces and the system generalized method of moments, Wang et al. [33] attested to an inverted U-formed relation between corporate environmental accountability and ecological policies. Meanwhile, Zhang et al. [34], in a finer way, exhibited that environmental regulations notably enhance corporate environmental responsibility through liability insurance. Collins et al. [35] analyzed a dataset of more than 25,000 manufacturing plants across the United States during 1998–2012, along with revealing that purposeful environmental approaches mitigate toxic discharges while preserving employment.

2.3. Studies on the Implementation of Difference-in-Differences Methodology

Among scholarly studies, the difference-in-differences model serves as an essential quasi-experimental technique for assessing policy impacts, with extensive applications across economic and ecological research [36,37,38]. Contemporary economic scholars have employed this approach to investigate diverse issues including income distribution disparities, industrial upgrading, labor market dynamics, and macroeconomic development. A representative application can be disclosed by Zhao et al. [12], which examined how technology–finance pilot programs influenced corporate environmental transparency. Their multi-phase DID analysis of China’s A-share listed firms (2008–2022) demonstrated significant policy effects in enhancing disclosure practices, with green credit mechanisms functioning as a moderating variable. Empirical studies employing DID methodology have yielded significant findings across various policy domains. Sendstad et al. [39] analyzed EU country-level data (2000–2017) through a DID framework, demonstrating that consistent policy frameworks with credible governmental commitments effectively stimulate private sector capital expenditures. By serving China’s National Broadband initiative as a natural experiment, Luo et al. [40] deeply inspected the causal nexus for digital infrastructure and entrepreneurial activity. Using a rigorous DID approach, the researchers demonstrated that the Broadband China policy implementation to some extent greatly boosted entrepreneurial vitality nationwide, serving as an impetus for business innovation and economic growth. Their findings provide compelling evidence that strategic digital infrastructure development can notably stimulate entrepreneurial ecosystems. In a finer way, Wu et al. [41] executed a comprehensive investigation into Chinese provincial-level (2006–2017) and municipal (2011–2018) data, disclosing that digital transformation greatly enhances energy efficiency and pollution control outcomes. Zhang et al. [42] employed a DID technique for accessing how carbon emission trading policies influenced the energy crisis. Their findings revealed that these measures not only alleviated energy shortages but also accelerated the conversion towards a cleaner, energetic mix in China. Ma et al. [43] engaged the PSM-DID technique to Chinese manufacturing firms (2009–2018), revealing that accelerated depreciation policies substantially boost both R&D expenditures and capital investments while curbing financialization tendencies. In environmental economics, the DID technique has also been highly valued. Wen et al. [44] studied the carbon mitigation through environmental tax policies through the PSM-DID model draw on a panel dataset of Chinese provinces covering 2004–2020. Their outcomes showed that implementing environmental taxes contributes to cutting carbon emissions. Contrasting results emerged from Nawaz’s [45] investigation of N-11 nations (2005–2019), where DID analysis indicated limited effectiveness of green finance initiatives in climate change mitigation. Moreover, academics have employed DID framework to explore interdisciplinary applications in healthcare [46]. Chen et al. [47] used a staggered DID technique to assess China’s urban–rural healthcare reforms spanning 2004–2020. Their findings demonstrated that this policy scheme effectively reduced household medical expenses while mitigating disparities in social welfare distribution.

In summary, a thorough review of the academia discloses that extensive research has been conducted on drivers of carbon emissions, economic and ecological impacts of environmental regulation, and advancements in the DID model, producing rich research results and countless valuable insights. Nonetheless, there are still four gaps in the existing literature. To be specific, firstly, detailed studies on measuring carbon emissions have been conducted, but there is a lack of unified understanding. Ignoring this issue puts policymakers in a situation of information asymmetry when implementing emission reduction measures, thereby impairing the fairness and transparency in carbon emissions. Secondly, while numerous academic studies are addressing determinants associated with carbon discharges, studies focusing specifically on green fiscal policy remain scarce. Without a focus on green tax policy, the formulation of effective fiscal means for carbon reduction and the establishment of coordinated regional carbon reduction mechanisms are significantly hindered. Thirdly, research exploring how EPT impacts carbon emissions is limited, especially empirical studies that systematically exhibit the effect mechanisms. Ignoring these discussions will make it difficult to fully undertake the pathways of incorporated technological innovation, industrial restructuring, and energy transition. Lastly, there is a lack of attention to the dynamic changes in EPT and the handling of endogeneity issues in the model. Failing to address these critical issues not only undermines the effectiveness of the model in evaluating policy outcomes but also affects policymakers’ decisions regarding subsequent policy adjustments. In conclusion, this study aims to utilize the DID method, the PSM-DID model, along with instrumental variable approach to evaluate policy effects. Parallel trend and placebo tests are conducted to verify the robustness of these impacts, providing valuable insights for policymakers aiming to reconcile environmental sustainability with economic growth.

3. Theoretical Analysis

3.1. Institutional Background of EPT

Light industry and heavy chemicals played a dominant role in Chinese industrial structure before the pre-reform era, which caused many ecological pollution problems. Out of expectations for environmental protection and sound progress, in a finer way, the Chinese government first proposed a pollution fee system in 1978. In addition, China further issued the Regulations on the Collection, Use, and Management of Pollution Fees in 2003, which marked a watershed for the transition of pollution charges from trial implementation to formal payment. Although pollution fees have achieved certain results in reducing pollutant emissions, due to the lack of legal enforcement support, problems such as incomplete collection scope, insufficient collection, and improper fund management have arisen in the actual implementation process. This not only disrupts the actual benefits for environmental regulations but also hinders the advance of ecological control technologies. However, amid the ongoing progress alongside China’s socioeconomic advancement, numerous previously unknown pollution factors have begun to emerge, increasing our pressure and responsibility for environmental protection. Therefore, since 2007, China has regarded the formulation of an environmental protection tax as an important task in tax and fee reform, aimed at addressing the challenges associated with enforcing environmental discharge charges and promoting policies tailored to national realities for China. The draft for Environmental Protection Law was reviewed and approved by Chinese National Congress in early 2016, which was further improved after multiple rounds of public solicitation of opinions and suggestions. At the end of 2017, Chinese Regulations for Environmental Protection Tax were officially passed and stipulated to be implemented nationwide from 1 January 2018, replacing nearly 40 years of pollution discharge fees. Relative to pollution charges, the imposition targets in environmental protection tax policy are more diverse, covering not only the conservation and management of key natural assets, including water systems, soil environments, and atmosphere, but also the regulation of contaminants, including acoustic pollution and solid refuse. Furthermore, the foundation of the ecological protection tax demonstrates greater scientific validity and rationality. Based on the actual number of taxable pollutants emitted by taxpayers, the pollution emissions of taxpayers are determined through measurement and monitoring. Once again, environmental protection tax has legal traceability, and taxpayers need to demonstrate as well as remit payments to the taxation authorities as mandated and accept supervision and inspection. This strengthens the supervision of environmental protection work and ensures its effective implementation. Finally, the environmental levy brings in multiple recently enacted levy-related incentives, including reductions and exemptions for qualifying clean-energy initiatives. To be more precise, the establishment of the environmental protection tax system constitutes China’s first green taxation framework and signifies a critical step toward advancing the country’s sustainable and low-carbon transition. Even so, it demonstrates Chinese government’s high attention and consistent adherence to the policy of integrating ecological civilization construction with economic and social development.

3.2. Theoretical Analysis and Research Hypothesis

To deduce how EPT affects carbon emissions from a mirco position, following Antweiler et al. [48], we build a corporate carbon emissions model that includes EPT, characterizes the economic behavior process of corporate carbon emissions, and derives the benefits for EPT on corporate carbon emissions reduction. Specifically, we consider the production process of a representative firm, where the firm needs to invest certain factors such as capital (labeled as k), data (labeled as d), and labor (labeled as l) into production, and assume that the Cobb–Douglas (simplified as C-D) production function is defined as follows:

$$y=A{k}^{\alpha }{d}^{\beta }{l}^{1-\alpha -\beta }$$

(1)

where y represents the output level. k, d, and l, in a finer way, reflect the inputs of capital, data, and labor, respectively. α and β denote the output elasticity of capital and data factors. A reflects the productivity level. Meanwhile, assuming that the firm’s production process generates carbon emissions, it can minimize these emissions by investing in carbon reduction equipment. More precisely, the firm’s carbon emissions scale and its output level are linearly related [49]. Thus, its scale can be explained as follows:

$$e=(1-\eta )\theta y$$

(2)

In Formula (2), e denotes carbon emissions scale. η signifies the investment level in emission reduction technology that effectively lowers carbon emissions, whereas θ signifies the carbon emission intensity per product unit (θ > 0). A firm always needs to pay an emission tax on each unit of carbon dioxide emitted during the production process. Assuming that the carbon emission tax is a quantity tax with a rate of T, and T > 0, the firm faces a total emission tax of Te. Meanwhile, when firms use emission reduction equipment for carbon reduction, they need to pay a certain amount of emission reduction costs, which are composed of variable costs and fixed costs. Referring to Xu and Song [50], we infer that the cost function of carbon reduction for firms is as follows:

$$c=a{\eta }^{b}y+f$$

(3)

In Equation (3), c reflects the firm’s carbon reduction cost. a reflects the variable cost coefficient. b reflects the cost elasticity of carbon reduction equipment investment, and a > 0, b > 1. f is the fixed cost of corporate carbon reduction. It can be seen that the first item on the right in Formula (3) exhibits the variable emission reduction cost, which is related to the production quantity of the firm’s products; and the second item represents the fixed emission reduction costs, which are not related to the quantity of products produced by the firm. We generally assume that firms can independently decide whether to minimize carbon dioxide generated during its production stage. If a firm does not reduce the carbon dioxide emissions, it does not need to pay the emission reduction cost c. In addition, in order to introduce EPT factors, the firm can obtain low-carbon transformation funding support from the local governments with a probability of λ, and 0 < λ < 1. We presume that a firm covers the expense for cutting carbon emissions in the production process through low-carbon financing, that is, a firm that receives low-carbon transformation funding is more inclined to restrain carbon dioxide emissions. Likewise, referring to Qi et al. [51], environmental regulatory authorities are introduced into the model. When the firm directly emits carbon dioxide without any carbon reduction treatment, there is a probability of (1 − m) that the firm will be punished by the environmental regulatory authorities. The ratio of fines to firm’s profits is δ, indicating that the probability of evading environmental penalties is m, and δ > 0, 0 < m < 1. Due to the negative behavior of firms directly discharging pollutants into the environment with a mentality of luck, when such behavior is discovered by environmental regulatory authorities, firms will face high punitive fines. Therefore, it is assumed that δ(1 − m) > 1. Therefore, assuming that the capital factor cost of the enterprise is r, the data factor cost is n, the labor factor cost is w, and the product pricing is p, the firm’s expected profit can be expressed as follows:

$$\pi =\lambda (py-\psi -Te-c)+m(1-\lambda )(py-\psi -Te)+(1-\lambda )(1-m)(1-\delta )(py-\psi -Te)$$

(4)

In Equation (4), ψ represents the total input cost of factors, i.e., ψ = (rk + nd + wl). In this context, firms make the decision to maximize profits by changing the investment intensity of carbon reduction equipment in the production process. By combining Equations (2)–(4), we compute the following first-order partial derivative in Formula (4) regarding η, arriving at the following result:

$${\displaystyle \frac{\partial \pi }{\partial \eta }}=T\theta y\left[\lambda +(1-\lambda )(1-\delta +m\delta )\right]-\lambda ab{\eta }^{b-1}y=0$$

(5)

By organizing Equation (5), the optimal investment intensity of firm’s carbon emission reduction equipment can be obtained as follows:

$${\eta }^{\ast }={\left\{{\displaystyle \frac{T\theta \left[\lambda +(1-\lambda )(1-\delta +m\delta )\right]}{\lambda ab}}\right\}}^{{\displaystyle \frac{1}{b-1}}}$$

(6)

Further combining Equation (2), the optimal carbon emission scale e* of the firm can be expressed as follows:

$$e^{\ast }=(1-{\eta }^{\ast })\theta y$$

(7)

To study how EPT influences the scale of carbon emissions for corporations, we start by calculating the partial derivative of Equation (7) relative to λ, employing a positive monotonic transformation of the function to derive the subsequent expression:

$${\displaystyle \frac{\partial e^{\ast }}{\partial \lambda }}={\displaystyle \frac{\partial e^{\ast }}{\partial {\eta }^{\ast }}}{\displaystyle \frac{\partial {\eta }^{\ast }}{\partial \lambda }}<0$$

(8)

The above formula means that EPT can effectively reduce carbon emission after providing low-carbon transformation financing support, thus proving the positive effect of EPT on corporate carbon reduction. Despite findings demonstrating that implementing EPT markedly helps in curbing carbon outputs, as shown in Equation (8), the specific mechanism by which EPT directly achieves this reduction still needs further elaboration. Therefore, we conduct theoretical analysis on this. Firstly, as a forceful tool for regulating carbon deluges, EPT notably promotes the emission reduction behavior of enterprises and individuals, enhances their cost- and energy-saving awareness, and stimulates corporate engagement in green and emission-reducing production modes as well as product designs through tax incentives, thereby supporting the dual carbon goal and sustainable development. In addition, EPT provides the government with fiscal revenue, supports environmental protection projects, promotes low-carbon economic transformation, promotes environmental technology innovation, assists cities in transitioning to low-carbon circular economy, accelerates green development, and achieves harmonious coexistence between economy and environment. These analyses further indicate that environmental protection taxes are beneficial for curbing carbon output. In summary, implementing environmental protection taxes curbs municipal carbon discharges while simultaneously fostering the green development of cities, achieving the goal of harmonious coexistence between the economy and the environment. Based on this, this study proposes assumptions as outlined below:

Hypothesis 1 (H1).

EPT policy can greatly lessen urban carbon emissions.

Through the above analysis, it has been demonstrated that EPT contributes to carbon discharge reductions in a direct manner. Further research has found that EPT can also indirectly promote carbon emissions reduction through the following channels.

The first channel is the energy utilization efficiency effect. To be specific, EPT imposes additional costs on energy consumption, particularly targeting high-pollution and high-energy-consuming activities. This cost pressure acts as a powerful incentive for firms to optimize their production processes and invest in energy-saving technologies. By reducing energy use, firms can lower their tax liabilities. This economic incentive encourages a wide range of industries to adopt more efficient practices. The result of these efforts is notable growth in the overall energetic performance of industries and urban operations. As energy use is optimized, carbon output per production unit of economic output is mitigated. Secondly, EPT typically includes tax incentives for renewable energy adoption coupled with green-tech breakthroughs. These incentives are designed to motivate companies to develop, apply, and use cleaner technologies, which further strengthen energy utilization while also alleviating attachment to fossil energy, a critical cause of carbon discharges. Adopting sustainable energy sources substantially lowers the city’s carbon footprint. Additionally, green innovation incentives motivate companies to allocate resources toward research and development aimed at technically innovative and environmentally friendly energy-saving methods. This continuous innovation is essential for long-term sustainability and can lead to breakthroughs that further reduce energy consumption and emissions. Lastly, EPT can drive structural changes in the economy by undermining the high-energy-dependent sectors’ competitiveness. As these industries face higher costs due to the tax, they become less competitive compared to more energy-efficient sectors. This cost pressure encourages a shift in economic activities towards sectors and fields that are more energy efficient. This reallocation of resources towards greener industries helps to form a more sustainable urban development model and reduces overall carbon emissions [19]. By incentivizing firms to become more energy-efficient and promoting renewable energy deployment, EPT migrates carbon discharges while fostering economic growth through improved technical innovation and energy efficiency. Following the preceding analysis, the following hypothesis is proposed:

Hypothesis 2 (H2).

EPT policy can greatly lessen urban carbon emissions by promoting energy utilization efficiency.

The second channel is the green technology innovation influence. To be specific, EPT increases carbon emissions costs. By levying taxes on polluting emissions, it directly increases the production costs of enterprises. In order to maintain profitability and competitiveness, enterprises are forced to seek ways to reduce costs. Green technological innovation provides an effective solution. For example, enterprises may invest in R&D in energy-saving technologies crucial for discharge reduction, such as advancing more efficient industrial boilers to cut energy usage alongside carbon output. This innovation not only helps enterprises meet environmental requirements but also brings long-term economic benefits by reducing operating costs [52]. Green taxation policy can stimulate innovation-promoting effects of the market. The implementation of EPT sends a clear signal to the market that environmental protection is of great importance. This encourages a large number of innovative companies and green tech innovation, alongside advancement within research institutions. Some start-up companies may focus on the development of sustainable energetic technologies, like wind-based alongside solar electricity generation, to provide cleaner energy alternatives. At the same time, the government may allocate green taxation income to support green technology innovation, such as providing incentives alongside subsidies for development alongside research projects or building public–private research and innovation platforms. These measures can further promote the development alongside adopting green technologies, hence mitigating urban carbon discharges. In general, the is crucial in advancing clean technical innovation as well as lowering municipal carbon discharges through the joint action of cost–pressure and market-stimulation mechanisms. Consequently, the research proposes the hypothesis outlined below:

Hypothesis 3 (H3).

EPT policy can notably lessen urban carbon emissions by enhancing green technology innovation.

This third channel concerns industrial structure upgrading effects. Regarding technical R&D, implementing EPT has increased environmental pollution expenditure costs, prompting enterprises to increase their R&D funding toward emissions mitigation, energy efficiency, low-carbon technologies, as well as other areas. At the same time, it encourages enterprises to apply advanced environmental protection technologies and equipment to achieve more efficient and cleaner production methods. R&D innovation not only demands substantial investment but also necessitates attracting high-end technology and innovative talent, facilitating the evolution of industrial composition featuring labor-based attributes towards tech-based sectors, achieving advanced alongside deeply unified production processes. Regarding government taxation, the domains addressing the environmental protection tax is wider, as well as the levying rate tends to exceed the contamination discharge fees. Environmental protection levy administration and oversight are stricter than those of pollutant emission charges. Under EPT framework, taxpayers must deliver relevant certification regarding ecological matters to back up their tax filings, and tax administrations have the ability to carry out traceability checks on them. When it comes to environmental protection charge, firms merely need to render in accordance with the stipulated time, and tax authorities are unable to decently care and regulate it [18]. Moreover, the Environmental Protection Agency has set up a more comprehensive reporting mechanism to encourage the public to actively engage in environmental supervision, promptly detect, and handle illegal and irregular behaviors. Consequently, the environmental tax collections scheme notably underlines tax control talents and high-standard supervision compels companies to cut down on non-competitive products and services, thus subtly propelling the industrial optimization and reform layouts. Without a doubt, industrial upgrading can lower carbon emissions coupled with attaining sustainable development goals by implementing advanced technologies, optimizing the industrial structure, and formulating corresponding policies. Grounded in the above, the presumption is addressed as follows:

Hypothesis 4 (H4).

EPT policy can greatly lessen urban carbon emissions by promoting industrial upgrading.

Carbon emissions, for example, exhibit a distinct trait of dynamic diffusion in space, which is greatly affected by adjacent areas. Thus, the spatial correlation among areas must be taken seriously. In addition, as economic actions among areas become steadily frequent, pilot cities may create demonstration and diffusion effects on adjacent areas, thereby affecting the efficacy for EPT on carbon emissions in neighboring cities. Firstly, pilot cities have achieved multiple goals of clean energy transformation and carbon reduction, which enhances the motivation of local officials to imitate low-carbon development in adjacent cities. Non-pilot cities can greatly reduce carbon emissions by learning and imitating the control experience and development models from pilot cities. Secondly, while promoting local low-carbon development, pilot cities have also to some extent broken down barriers for factors flow, enhanced the connections within low-carbon industries and other industries, and promoted the exchanges between surrounding areas and pilot cities. Economic exchanges among cities contribute to the diffusion of low-carbon knowledge and technology, offering the experience for surrounding areas in emission reduction, thereby promoting the coordinated control of carbon emissions in adjacent areas. Then, we propose the following hypotheses:

Hypothesis 5 (H5).

EPT policy has spatial spillover effects, which have a suppressive effect on carbon emissions in neighboring areas.

In brief, following the above analysis, as depicted in Figure 1, the theoretical mechanisms by which EPT affects carbon emissions can be briefly disclosed as follows.

4. Materials and Methodology

4.1. The Specification for Benchmark Panel Econometric Model

This study mainly deliberates how EPT affects carbon emissions. Even so, given that the policy has been implemented synchronously in all cities across the country since 2018, using traditional DID models for estimation may lead to regression bias. Therefore, referring to Xu and Huang [53], as claimed by the median of 2018 pollution tax collection criteria at the province level, we split the entire sample into control unit and trial unit. That being the case, as claimed by Model (9), an intensity difference-in-differences (I-DID) technique is subtly employed to disclose the benefit of EPT on carbon emissions.

PCE_it = α + β × Tax_did_it + γ × Control_it + μ_t + λ_i + ε_it

(9)

where i and t reflect city and year, respectively. As listed in Model (9), all variables are demonstrated as follows: the explained variable PCE_it reflects per capita carbon emissions upon an urban scale. The key explanatory variable for EPT is denoted by Tax_did_it, which is formulated by multiplying policy indicator and time indicator. Most importantly, Control_it exhibits the control variables at the prefecture scale, in a finer way, including economic development level (PGDP), population density (Popul), financial development (Finance), science and education spending (Sciedu), industrialization degree (Indus), and fiscal revenue (Revenue). In a finer way, α, β, and γ denote regression coefficients, as claimed in Model (9). To be more precise, β exhibits how EPT affects carbon emissions at the city level. That is, if β < 0, it reveals that EPT to some extent can lessen urban carbon emissions. Otherwise, EPT will aggravate urban carbon emissions. In addition, μ_t and λ_i represent the time fixed effects (Time FE) coupled with the individual fixed effects (ID FE). Even more, ε_it represents the random error term.

4.2. Variable Selection and Description

4.2.1. Explained Variable: Per Capita Carbon Emissions (PCE)

Regarding the studies on carbon emissions at the areal scale, the critical facet exhibits how to attain carbon emission data, notably for developing nations [54]. It is also becoming gently urgent following that China has not yet officially demonstrated the city-level carbon emission. That being the case, scholars have essentially utilized three techniques to quantify Chinese carbon emissions at the city scale: high-resolution spatial gridding technique reflecting spatial emission patterns, nighttime light data approach using satellite observations, and energy consumption method rooted in fossil fuel usage [55,56,57]. To ensure data consistency and completeness, we employ the energy consumption method to attain carbon emissions at the city level, covering four primary energy categories. In a finer way, the related formula is detailed below, as presented in Equation (10).

CE = CE_h + CE_e + CE_p + CE_g = χ × E_h + φ × E_e + ν × E_p + κ × E_g

(10)

where CE, as demonstrated by Equation (2), reflects total urban carbon emissions. Even more, CE_h, CE_e, CE_p, along with CE_g correspond to emissions derived from four energy sources: heating (h), electricity consumption (e), liquefied petroleum gas (p), and natural gas (g), respectively. Eh, Ee, Ep, and Eg denote the consumption for these four energy sources spanning 2006–2022, respectively. χ, φ, ν, and κ are carbon conversion coefficients for raw coal, electricity, liquefied petroleum gas, and natural gas, respectively. In line with IPCC criteria, emission coefficients are demonstrated as 1.9003 kg CO₂/kg for raw coal, 3.1013 kg CO₂/m³ for liquefied petroleum gas, and 2.1622 kg CO₂/m³ for natural gas. Carbon emissions stemming from electricity consumption are difficult to assess, particularly since 2011, when China segmented its power grid into six areas: eastern, northern, central, southern (Hainan merged into the southern grid), northeastern, as well as northwestern. Therefore, we subtly reckon carbon emissions induced by electricity use in each region based on the historical CO₂ emission baseline in China. Thus, to fully disclose urban carbon distribution from a geographical perspective, we select the amount of carbon emissions at the prefecture level relative to its registered residence population over the years (PCE) as the explained variable (as shown in Figure 2). Furthermore, to strengthen the validity of practical findings, we also fulfill robustness testing using carbon emission intensity.

4.2.2. Key Explanatory Variable: Environmental Protection Tax (EPT)

It should be noted that EPT is put into action in whole China since 2018; that being the case, this law must be strictly observed for all cities without exception. In this way, if traditional DID models are used, for the whole sample, it will be hard to discern the control group along with the treatment group, resulting in the inability to disclose the implementation impacts of EPT. Hence, referring to Xu and Huang [53], we use an intensity DID (I-DID) technique to assess the benefits for EPT on carbon emissions at the city scale. First, as claimed by the median of 2018 pollution tax revenue at the province level, we set the policy dummy (PD) variable. To be specific, a city is deemed to be in the trial group if its pollution tax rate exceeds the median; in that case, the value for its PD variable equals to 1. Otherwise, this city will be integrated into the control group, and the value for its PD variable is 0 (as depicted in the left of Figure 3). Second, viewing the layout of time dummy (TD) variable for EPT, 2018 being the baseline line, the value of TD variable for 2018 followed by later periods equals to 1. Otherwise, the value for its TD variable is 0. Even more, we form the interaction term between the PD variable and the TD variable at the city level, which reflects the key explanatory variable (Tax_did) listed in Model (9). In addition, all cities have collected pollutant emission fees before 2018. With the wide implementation of EPT, some cities have raised their standards, while others have kept their pollution levies at their initial levels. Therefore, this study recreates each city’s PD variable for robustness testing, as attested by the case of tax burden transfer. To be more precise, after 2018, the value of a city’s PD variable equals 1 if its pollution tax rate rises. Otherwise, the value for its PD variable is 0 (as illustrated in the right of Figure 3). On this basis, to further obtain the indicator for EPT in the Model (9) under tax burden transfers, we multiply PD variable by TD variable for each city eventually.

4.2.3. Control Variables

What needs to be demonstrated is that carbon emissions may also be impacted by variables like economic conditions or change in population at the city level, aside from the impact of EPT. To fully disclose all potential sources and mitigate estimation errors, we add a number of control variables into Model (9), such as financial development (Finance), industrialization degree (Indus), science and education level (Sciedu), economic development (PGDP), fiscal revenue (Revenue), and population density (Popul). To be specific, financial development (Finance), all the more so, is demonstrated by the share of financial entities’ year-end lending amounts to GDP at the city level. The degree of industrialization (Indus), to a greater extent, is demonstrated by the share of secondary output at the city level. The level of science and education (Sciedu), as noted in Model (9), is demonstrated by the share of scientific and educational expenditures to GDP at the city level. Economic development (PGDP), which is desirably obtained by deflating the nominal per capita GDP utilizing Consumer Price Index (CPI) at the area level, is reflected as each city’s real per capita GDP. fiscal revenue (Revenue) is notably demonstrated by the degree of fiscal revenue to local GDP at the city level. Meanwhile, the symbol Popul, subtly demonstrated by permanent residents to its areas at the city level, exhibits population density.

4.3. Data Source and Processing

As is well known, empirical evaluation implies data similarity, availability, and integrity. That being the case, to attain a novel dataset spanning 2006–2022, we meticulously align the EPT scheme with prefecture-level data claimed by Model (9). In this background, we ultimately retrieve a directory embracing 282 cities from 29 Chinese provinces, given that certain cities exhibit missing data. To be more precise, the dataset used for empirical analysis, attested by Model (9), is retrieved from China Statistics Yearbooks at the area level, China Urban Statistics Yearbooks at the city level, and EPS Database. More importantly, by retrieving the website of the environmental protection bureau at the province level, the data for each city’s pollution fee collection criteria is formed. In a finer way, by crawling such yearbooks at the province or city level as China Energy Yearbooks at the province level, coupled with China Urban–Rural Construction Yearbooks at the city level, the essential dataset is formed to figure out carbon emissions at the city level eventually. In particular, employing such techniques as ARIMA coupled with linear interpolation, we have patched certain data that is unavailable. Essentially, to mitigate heteroscedasticity from the statistical view, all regular variables, at the 1% quantile, are subjected to Winsor 2 truncation, and all absolute variables have been logarithmized. As depicted in

Table 1. Descriptive statistics.

Variable	Definition	The Entire Sample				Pilot Cities		Non-Pilot Cities
		Obs	Mean	S.D.	VIF	Mean	S.D.	Mean	S.D.
PCE	Urban carbon emissions	4794	1.074	0.701	—	1.009	0.642	1.154	0.759
Tax_did	EPT scheme	4794	0.162	0.368	1.32	0.294	0.456	0	0
Pgdp	Economic development	4794	10.529	0.722	2.00	10.542	0.704	10.514	0.743
Indus	The degree of industrialization	4794	0.378	0.078	1.75	0.382	0.069	0.374	0.087
Finance	Financial development	4794	0.643	0.254	1.99	0.606	0.242	0.689	0.261
Revenue	Fiscal revenue	4794	0.069	0.025	1.60	0.067	0.026	0.071	0.024
Sciedu	Science and education level	4794	0.034	0.017	1.59	0.032	0.012	0.037	0.021
Popul	Population density	4794	5.738	0.925	1.18	6.129	0.644	5.262	0.990

Note: Obs denotes the number of observed values. Mean and S.D. reflect the mean value and standard deviation for each variable. VIF denotes variance inflation factor. If the value is less than 10, it discloses that multicollinearity can be rejected in the model.

, we yield 4794 observations for empirical testing.

5. Empirical Results and Analysis

5.1. Baseline Regression Analysis

As mentioned in theoretical analysis, EPT can greatly reduce carbon emissions at the city level. To prove this hypothesis, we use fixed effects technique demonstrated in Model (9) to disclose how EPT affects carbon emissions at a city scale through gradually incorporating control variables. Notably,

Table 2. The baseline estimation outcomes for EPT on carbon emissions at the prefecture level.

Variable	PCE				CE
	(1)	(2)	(3)	(4)	(5)
Tax_did	−0.103 ***	−0.111 ***	−0.113 ***	−0.109 ***	−0.178 ***
	(−7.54)	(−8.14)	(−8.28)	(−7.85)	(−7.39)
Pgdp		0.108 ***	0.097 ***	0.097 ***	0.756 ***
		(4.74)	(4.11)	(4.11)	(18.49)
Indus		−0.125	−0.177	−0.160	−0.238
		(−1.12)	(−1.59)	(−1.42)	(−1.22)
Finance		−0.117 ***	−0.129 ***	−0.130 ***	−0.061
		(−3.18)	(−3.45)	(−3.46)	(−0.94)
Revenue			1.670 ***	1.638 ***	1.490 ***
			(5.45)	(5.34)	(2.80)
Sciedu			−2.714 ***	−2.604 ***	0.933
			(−3.92)	(−3.74)	(0.77)
Popul				−0.105 *	0.418 ***
				(−1.70)	(3.90)
_Cons	1.091 ***	0.078	0.201	0.794 *	−4.031 ***
	(286.49)	(0.33)	(0.80)	(1.85)	(−5.43)
ID FE	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes
R2	0.906	0.907	0.908	0.908	0.904
F statistic	56.86	27.29	24.42	21.36	77.00
Obs	4794	4794	4794	4794	4794

Note: t value is enclosed in parenthesis. In a finer way, at the 10% and 1% levels, significances are reflected by * and ***.

discloses the findings of estimation.

More specifically, as stated in Table 2, the effect of EPT on carbon emissions is analyzed at the city level. Firstly, without being affected by control factors displayed in Column (1), the value for EPT persists as negative at the 1% level, disclosing that EPT significantly inhibits urban carbon emissions. In addition, control variables, City FE, and Year FE are sequentially added, as presented in Columns (2)–(4), where the coefficients for EPT in all models remain greatly negative. To be more specific, taking Column (4) as an instance, the estimated value for EPT, at the 1% level, preserves as −0.109 after adding all control variables, indicating that a surge of one unit of standard deviation in EPT can result in a decrease of 0.057 (equals to −0.109 × 0.368/0.701) units of standard deviation in carbon emissions. That being the case, EPT can greatly lessen carbon emissions at the area scale, thus confirming the Hypothesis 1 (H1). Furthermore, even if the dependent variable demonstrated in Model (9) is changed from per capita carbon emissions at the area scale (PCE) to total carbon emissions (CE), with the result exhibited in Column (5), the value for EPT still remains greatly negative, further demonstrating the accuracy of the baseline estimation findings, which is also confirmed by the findings from Collins et al. [35].

Next, to disclose how control variables affect carbon emissions at the area scale, we analyze the estimated coefficient for each variable separately. From economic development at the city level (Pgdp), the results for its benefit upon carbon emissions remain considerably active in all models. By way of illustration, the estimated result for economic development on PCE is 0.097 at the 1% significance level, as shown in Column (4), implying to a greater extent that a rise in economic development notably results in a surge in energy usage, thus making carbon emissions control more difficult. The coefficients for financial development (Finance), science and education spending (Sciedu), and population density (Popul) remain both greatly negative, disclosing that the improvement for financial development, science and education spending, and population agglomeration to a certain degree can lessen urban carbon emissions. However, the coefficient for fiscal revenue (Revenue) remains notably positive, meaning that the large investment of fiscal revenue in high-energy production projects increases energy usage, thus yielding a lift in carbon emissions at the city level. Even more, the coefficients for industrialization (Indus) are not significant, as is plainly evident, indicating that the bonus for industrialization on carbon emissions at the city level is limited. Therefore, we can reduce urban carbon emissions by taking measures such as promoting financial development, expanding science and education spending, and stimulating population agglomeration.

In brief, the above findings reveal that EPT, financial development, science and education spending, and population agglomeration greatly reduce urban carbon emissions. On the contrary, economic development and fiscal revenue greatly lead to a sustained expanding for carbon emissions at the city level.

5.2. Parallel Trend Testing and Placebo Testing

In a general way, to enhance the validity of the benchmark regression outcomes exhibited earlier, we further perform the investigation by adopting parallel trend testing as well as placebo testing, as disclosed in Figure 4 and Figure 5.

First, the parallel trend assumption, in a finer way, discloses a reputed necessity for using the DID technique. In other words, the treatment group and the control group in many ways exhibit a common trend before policy intervention. Therefore, to fully observe whether the treatment group and the control group disclose the same trend before EPT execution, this study conducts a parallel trend check for EPT. To be more precise, following Beck et al. [58], as shown in Figure 4, we adopt two measurement methods to separate the treatment group and the control group, namely the median of environmental tax ratio at the area level and whether to adjust the collection standards after EPT execution. As depicted in Figure 4, regardless of the classification method adopted, there are no notable differences in the coefficients for EPT before policy execution, indicating that the treatment group and the control group, to a greater extent, exhibit a common tendency. However, after the implementation of EPT, the policy coefficients notably deviate from zero, as being plainly evident, implying a critical difference. Therefore, the parallel trend assumption is verified.

Second, to alleviate estimation bias originating from random factors in the model, we further validate its effectiveness using placebo testing. To be specific, we randomly select a certain number of cities from the entire sample as the pseudo trial group, with the extra cities being the pseudo control group, to estimate the pseudo sample and compare the distinction from the pseudo policy impact along with the real impact. If the results of the two are notably different, the benchmark regression results are valid. Otherwise, their results are invalid. Meanwhile, the study uses PCE and CE as the dependent variables to create 1000 pseudo experimental group for placebo testing, separately. As depicted in Figure 5, where the real treatment effect greatly deviates from those for the pseudo experimental group, and most of the coefficients for the pseudo experimental group are not significant. Based on this, the benchmark regression results are further validated.

Moreover, the latest DID research proves that pre-processing trend tests cannot serve as evidence for the validity of parallel trend testing and may even cause the deviations of estimated results and inferences [59]. Then, Rambachan and Roth [60] proposed a counter-factual technique for parallel trend testing, whose core idea is to conduct sensitivity analysis on point estimators and confidence intervals. To be more precise, it is necessary to build the maximum degree of deviation (Mbar) where parallel trends are violated, and the confidence interval of the processed point estimate corresponding to the aforementioned degree of deviation. If the confidence interval for processed point estimate excludes a zero value at the maximum deviation, it indicates that the treatment effect greatly exhibits robustness relative to the deviation of parallel trends. Therefore, referring to Biasi and Sarsons [61], the study sets Mbar (equals to 1 × standard error) to test the parallel trend sensitivity of the post-treatment effect for the EPT scheme, as shown in Figure 6.

Figure 6 shows the sensitivity test results of the parallel trend hypothesis for the year of policy implementation under relative deviation and smoothing constraints. The results show that, under the relative deviation constraint, the carbon reduction effect for EPT implementation in the current year is very robust. Similarly, under the smoothing constraint, the carbon reduction effect for EPT implementation in the year still remains robust. The above test results disclose that even if there is a certain degree of deviation from the parallel trend, EPT execution still greatly inhibits carbon emissions on a city scale.

5.3. Robustness Analysis

To eliminate possible concerns in the data, models, and results used in the research, and to enhance the credibility for policy treatment effect, we further use eight techniques to illustrate the robustness for the EPT scheme mentioned in the baseline estimation. That being so, the methods and results are disclosed as follows, as noted in

Table 3. The robustness testing results for EPT on carbon emissions at the city level.

Variable	CE_GDP	Adjust DID	Adjust Sample	Adjust Interval	PSM-DID	DML	Lag Control	IV Estimation
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Tax_did	−0.065 ***	−0.131 ***	−0.108 **	−0.089 **	−0.109 ***	−0.127 ***	−0.109 ***	−0.110 **
	(−2.74)	(−3.28)	(−2.45)	(−2.38)	(−2.63)	(−7.89)	(−2.83)	(−2.52)
_Cons	0.858	0.568	−1.748	0.878	0.794	0.000	0.411	—
	(1.19)	(0.36)	(−1.29)	(0.54)	(0.50)	(0.07)	(0.24)	—
Control	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
ID FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.752	0.909	0.905	0.919	0.908	—	0.907	—
F statistic	3.48	4.00	4.03	2.40	4.11	—	4.19	4.07
Obs	4794	4794	4199	3102	4794	4794	4512	4794
K.P.-rk-LM								251.267
C.D.-Wald-F								1.3 × 104

Note: t value included in parentheses. In a finer way, at the 5% and 1% levels, significances are reflected by ** and ***.

.

First, replace the dependent variable. To the best of our knowledge, the model may create different results due to different measurement methods for the dependent variable, easily yielding estimation bias. Accordingly, we use the GDP to carbon emissions ratio (CE_GDP) at the city level as the dependent variable for regression analysis. To be specific, as stated in Column (1), the findings for EPT, at the 1% significance level, remain greatly negative, to a greater extent implying that the robustness test preserves as sound through replacing the dependent variable.

Second, change the method of determining pilot zones. To greatly weaken the estimation bias originated from the division criteria between the control group and the treatment group, this study further uses the adjustment of pollution tax collection rate at the city level after policy implementation as the indicator to re-divide the control group combined with the treatment group. To be more precise, cities that have increased their original collection level are set as the treatment group, on the contrary, those that have not are set as the control group. That being the case, the product of the year dummy indicator served as the proxy of EPT execution for re-estimation. As exhibited in Column (2), the coefficient for EPT plan serves −0.131 at the 1% level, to a greater extent confirming the robustness of the results.

Third, adjust the regression sample. Directly administered municipalities, sub-provincial areas, as well as capital cities at the province level have greater autonomy in policy formulation and implementation, and they usually have more resources, policy incentives, etc. These peculiarities may lead to biased estimation results. That being the case, to better evaluate the applicability of the model and the robustness for EPT, this study excludes these cities and conducts regression analysis again. From a statistical view, as stated in Column (3), the value for EPT plan greatly serves −0.108 at the 5% level, to some extent implying that the robustness of benchmark outcomes disclosed above is further verified.

Fourth, adjust the sample interval. In general, the regression results will be affected by the occurrence of extreme events. Since the 2008 financial disaster and the COVID-19 epidemic may greatly reduce the benefits of the EPT scheme on carbon emissions, we remove the samples from 2006 to 2008 and 2020–2022, and further conduct regression testing. To be more precise, the coefficient for EPT, as stated in Column (4), also remains −0.089 at the 5% level, to some extent disclosing that the benchmark conclusions are robust.

Fifth, adjust the estimation method. The study combines propensity score matching (PSM) technique with DID model to jointly evaluate how the EPT scheme affects urban carbon emissions. To be specific, using radius matching technique to match the treatment group (TB) with a close control group (CB), the technique fully reduces the differences in observable variables between TB and CB, virtually solving the sample self-choice bias, to some extent improving the accuracy and credibility of EPT effect estimation. The coefficient for EPT is −0.109, with the 1% significance level disclosed in Column (5), further indicating the reliability of the benchmark conclusions disclosed by Model (9).

Sixth, use a dual machine learning (DML) technique to avoid estimation errors. The DID model did not cluster and summarize the average treatment effects of all individuals when calculating the overall average treatment effect, which may lead to biased estimation results. Referring to Zhang et al. [62], the study divides the dataset into seven subsets and uses the Lasso algorithm to treat nonlinear data, avoiding model errors and ensuring the unbiasedness of coefficient estimates, as noted in Column (6). To be precise, the estimated value for EPT, at the 1% level, remains greatly negative, verifying that the negative nexus between environmental tax reform and carbon emissions still holds at a city scale.

Seventh, perform first-order lag treatment on the control variables. To greatly mitigate endogeneity and improve estimation accuracy, this study lags all control variables detailed in Equation (9) by one period to lessen the reverse causality relation between EPT and carbon emissions at a city scale over the period 2006–2022. After expelling the lagged effects from control variables, as disclosed in Column (7), the coefficient for the EPT scheme on carbon emissions, at the 1% level, remains negative, to some extent confirming the robustness of the benchmark findings.

Finally, adopt lagged air circulation coefficient as the instrumental variable (IV). As plainly evident, when the amount of pollutant emissions is constant, the detection value of pollution concentration is inversely related to air spread attribute. Theoretically, the smaller the air spread value, the stronger the environmental tax collection standard will be, satisfying the correlation assumption. Most importantly, the lagged air circulation coefficient, as a natural factor, exhibits little impact on the current carbon emissions, which meets the exogeneity assumption. Thus, we address potential endogeneity exposed in the model by constructing IV, using the interaction term obtained by multiplying the natural logarithm of lagged air circulation coefficient with the time dummy variable at a city scale. At the 5% level, the value for Tax_did, as depicted in Column (8), serves −0.110 after using IV technique, confirming the reliability of the benchmark results. Meanwhile, Cragg–Donald Wald F statistic (C.D.-Wald-F), coupled with Kleibergen–Paap rk LM statistic (K.P.-rk-LM), both remain greater than 10, verifying the correlation assumption and the exogeneity assumption well.

5.4. Policy Uniqueness Test

As part of the study process, various facets are interrelated and influence each other, which can easily lead to confusion and interference, and the execution of pilot policies is no exception. That being the case, we further validate the benefits for EPT on carbon reduction by considering the impact of the execution of other pilot policies spanning 2006–2022. To be more precise, four policies related to local environmental governance are further integrated into the model. For example, in 2012, the New Energy Demonstration City (NewE) scheme was initiated. In 2017, the Green Finance Reform Pilot Zones (GFR) plan was performed. In 2010, the Low-carbon City Pilot (LowC) was subtly implemented. In 2011, the Energy Saving and Emission Reduction Fiscal Policy (ES&ER) was put into action. Thus, to greatly disentangle the impacts of these policies on EPT, as noted in

Table 4. The policy uniqueness testing results for EPT on carbon emissions at the city level.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
Tax_did	−0.108 ***	−0.111 ***	−0.111 ***	−0.107 ***	−0.110 ***	−0.187 ***
	(−2.61)	(−2.67)	(−2.66)	(−2.60)	(−2.63)	(−2.70)
Energy_did	−0.031				−0.035	−0.139 **
	(−0.82)				(−0.94)	(−2.22)
Green_did		−0.087			−0.084	−0.357 ***
		(−1.08)			(−1.09)	(−3.74)
Low_did			−0.025		−0.018	−0.088 *
			(−0.85)		(−0.62)	(−1.88)
Fiscal_did				−0.122 **	−0.120 **	−0.221 ***
				(−2.50)	(−2.45)	(−2.77)
_Cons	0.747	0.707	0.743	0.810	0.637	−4.739 **
	(0.48)	(0.44)	(0.47)	(0.53)	(0.42)	(−2.48)
Control	Yes	Yes	Yes	Yes	Yes	Yes
ID FE	Yes	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.908	0.908	0.908	0.909	0.909	0.906
F statistic	3.75	3.71	3.72	4.01	3.28	10.71
Obs	4794	4794	4794	4794	4794	4794

Note: t value is enclosed in parenthesis. In a finer way, at the 10%, 5%, and 1% levels, significances are reflected by *, **, and ***.

, this study gradually input them into Model (9) for empirical testing.

As stated in Table 4, the estimated results for four pilot policies incorporated into Model (9) are reported. By introducing these different policies mentioned above, the outcomes of policy uniqueness testing, to a greater extent, proves that EPT substantially reduces urban carbon emissions during 2006–2022. To be specific, we first inject the pilot plan of NewE (Energy_did) into Model (9). To be presented in Column (1), the value for EPT on carbon emissions remains negative, from a statistical point of view, with an estimated value of −0.108 and a significance level of 1%. Further, the pilot plan for GFR (Green_did), the pilot plan for LowC (Low_did), as well as the pilot plan for ES&ER (Fiscal_did) are sequentially added to the model for empirical testing, as presented in Columns (2)–(4), respectively. As plainly evident, the results for EPT significantly remain negative. In addition, when four policies are jointly added to the model, as demonstrated in Column (5), the estimated value for EPT is −0.110 at the 1% level. Finally, as shown in Column (6), when the dependent variable is replaced with total carbon emissions and four policies are simultaneously introduced, the regression results remain significantly negative. It reveals that the benefits of the EPT scheme on carbon emissions are, to some extent, not affected by those pilot plans, inevitably validating the cogency for the benchmark results demonstrated above.

5.5. Heterogeneity Analysis

To best our knowledge, the findings of the basic regression, in a finer way, prove that EPT substantially reduces urban carbon emissions over 2006–2022. However, China spans a big area, exhibiting critical disparities in location, resource endowments, and policy environments across cities. Most importantly, the effectiveness of EPT varies among different cities due to their own and external conditions. For example, cities with higher levels of innovation are often able to adopt highly developed technologies aiming at energy-saving and emission reduction. Therefore, cities with abundant fossil energy resource endowments have relatively high carbon emissions. Then, the differences in industrial structure also exhibit power upon carbon emissions. Therefore, to implement policies more adequately according to different types of cities, this study conducts heterogeneity testing based on geographical location, innovation degree, industrialization scale, and resource endowment. The results are disclosed in

Table 5. Heterogeneity testing results for EPT on carbon emissions at the city scale.

Variable	Geographical Location		Innovation Attribute		Resource Attribute		Industrial Attribute
	East	Others	Inno	Non-Inno	Non-Res	Res	Non-Old	Old-Ind
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Tax_did	−0.001	−0.169 ***	−0.152 **	−0.113 **	−0.182 ***	−0.046	−0.155 ***	−0.030
	(−0.02)	(−3.54)	(−2.42)	(−2.28)	(−3.42)	(−0.66)	(−2.84)	(−0.55)
_Cons	5.235 ***	−2.174 *	5.746 ***	−3.106 **	5.725 ***	−5.559 ***	5.338 ***	−3.465 ***
	(2.88)	(−1.89)	(2.61)	(−2.30)	(3.00)	(−3.82)	(2.67)	(−2.71)
Control	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
ID FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.912	0.909	0.921	0.895	0.917	0.908	0.904	0.938
F statistic	11.63	7.34	6.13	3.63	3.71	5.49	4.84	3.31
Obs	1700	3094	1275	3519	2856	1938	3179	1615

Note: t value included in parentheses. In a finer way, at the 10%, 5%, and 1% levels, significances are reflected by *, **, and ***.

.

Firstly, from the geographical location, the sample is split into eastern cities (East) as well as central and western cities (Others). As noted in Table 5, Columns (1) to (2), to some extent, disclose the test findings for East and Others, respectively. As a whole, EPT can greatly lessen carbon emissions in Others, but is not significant in East. In theory, whether EPT can restrain carbon emissions at a city scale notably depends on whether it can promote industrial upgrading and green technology innovation. Eastern cities have shifted their pollution intensive industries to Others earlier, resulting in a pivotal depletion in carbon emissions. Therefore, EPT’s impact on carbon emissions is relatively trivial. However, the industrial structure in Others represents heavy industry, and the industrial transfer is relatively slow. Unlike the East, industrial transfer mainly focuses on external transfer. The industrial transfer in Others is mainly based on the internal transfer of comparative advantages, and the quantity of carbon emissions in Others is relatively large during the sample period. Therefore, the carbon reduction benefits from green innovation and industrial upgrading initiated by the EPT scheme are notable.

Secondly, we split the full sample into innovative cities (Inno) as well as non-innovative cities (Non-Inno). To be more specific, Columns (3) and (4) disclose the empirical test findings for Inno and Non-Inno, respectively. On the whole, EPT can more effectively lessen carbon emissions in innovative cities, but the impact in non-innovative cities is poor. A plausible claim is that innovative cities, in a sense, exhibit more advanced technologies and a stronger capacity for green innovation. This enables them to respond more quickly to EPT by adopting cleaner production and renewable energy, realizing the green transition of industrial composition, thus achieving greater emission reductions. However, non-innovative cities may lack the vital technological infrastructure and innovation capabilities to make critical changes in response to EPT. Thus, they may struggle to adopt effective means to reduce carbon emissions, leading to a less pronounced reduction effect. More importantly, the industrial structure for non-innovative cities might be more reliant on traditional and high-emission industries, which are harder to transform quickly.

Thirdly, from the resource endowment, cities are divided into two types, as it were, covering resource-based areas (Res) along with non-resource areas (Non-Res), and the regression analysis is conducted separately. To be precise, as depicted in Columns (5) to (6), the value for EPT in non-resource cities remains notably negative. Conversely, the value for resource-based cities is not significant. The main reason lies in the differences in industrial structure, economic models, and energy consumption structure. Non-resource-based cities to some extent exhibit different industrial structures and can flexibly adjust their economic models and energy consumption structures in response to environmental tax policies, thereby reducing carbon emissions. However, cities that rely heavily on natural resources remain resource-based. Faced with economic development, resource-based cities may easily form a growth paradigm depicted by high energy usage and high pollution discharges. That being the case, environmental control technology and funds are fairly insufficient, resulting in limited emission reduction effects of EPT.

Finally, from the industrial attribute, we split the full sample into old industrial base areas (Old-Ind) coupled with non-old industrial base areas (Non-Old), and the regression analysis is conducted separately. To be specific, at the 1% level, as exhibited in Columns (7) to (8), the result for EPT remains greatly negative, implying that EPT notably lessens carbon emissions in Non-Old, but the coefficient for EPT in Old-Ind is insignificant. The reason, in a finer way, is concluded that non-old industrial base cities exhibit a more modern and differentiated industrial structure, which can quickly adapt to environmental tax policies, promote industrial upgrading and green transformation, thus reducing carbon emissions at the city level. However, old industrial base cities retain outdated and high-emission industrial systems, making it difficult to implement green innovation and industrial upgrading, and their sensitivity to EPT is relatively low. In addition, there may be deficiencies in the format and operation of environmental protection facilities in old industrial bases, which limits their emission reduction effects.

5.6. Transmission Mechanism Analysis

Following the basic regression outcomes listed above, it has been confirmed that EPT to some extent has effectively lessened urban carbon emissions. However, what are the mechanisms by which EPT at the area level affects carbon emissions? Therefore, following Xu et al. [63] and combining the aforementioned theoretical analysis, we conduct empirical testing, as shown in

Table 6. The mechanism testing results for EPT on carbon emissions at the city level.

Variable	Energy Utilization Efficiency			Green Technological Innovation			Industrial Structure Upgrading
	Energy	Energy	PCE	Patent	Patent	PCE	Ind_up	Ind_up	PCE
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Tax_did	0.183 ***	0.199 ***		0.142 ***	0.182 ***		0.013 ***	0.016 ***
	(7.34)	(7.98)		(3.81)	(4.88)		(3.70)	(4.62)
Tax_did_high			−0.201 ***			−0.086 ***			−0.089 ***
			(−4.28)			(−5.86)			(−6.16)
Tax_did_low			0.123 ***			−0.021			0.002
			(7.71)			(−0.62)			(0.05)
_Cons	2.939 ***	3.175 ***	1.297 ***	−7.888 ***	−7.518 ***	0.826 **	5.199 ***	5.230 ***	0.668
	(3.82)	(4.11)	(3.21)	(−6.86)	(−6.52)	(2.02)	(47.39)	(47.49)	(1.62)
Control	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
ID FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.719	0.720	0.641	0.898	0.899	0.637	0.955	0.955	0.638
F statistic	15.04	16.46	334.14	23.24	24.61	328.62	2197.84	2202.68	329.07
Wald test			44.47			3.20			4.56
Obs	4794	4794	4794	4794	4794	4794	4794	4794	4794

Note: t value included in parentheses. In a finer way, at the 5% and 1% levels, significances are reflected by ** and ***.

, from the three channels of energy utilization efficiency (Abbreviated as Energy), green technology innovation (Patent), and industrial upgrading (Ind_up). To be more precise, we specify energy intensity as the measure of energy utilization efficiency, the quantity of green inventions at a city scale realized in this period as a sign for green technological innovation, and the tertiary industry output at the city level to its GDP share as the indicator for industrial upgrading. In addition, the study uses two methods to construct the EPT interaction term and studies its impact on energy utilization efficiency, technological innovation, and industrial upgrading, respectively. To be more precise, firstly, the control group and experimental group are divided relying on the environmental tax collection standards, and the findings are disclosed in Columns (1), (4), and (7). Next, the control unit and experimental unit are split based on tax burden transfer, and the outcomes are exhibited in Columns (2), (5), and (8). Furthermore, to demonstrate the robustness of impact mechanisms, we further reclassify the EPT variable via each mechanism variable’s percentile of below 30% (low group) and above 70% (high group), and detect how EPT affects carbon emissions on an urban scale. That being the case, the results for EPT, as disclosed in Columns (3), (6), and (9), subtly confirm the robustness for energy utilization efficiency effect, enhancing green technological innovation effect, and industrial upgrading effect.

The regression findings of mechanism testing for EPT upon carbon emissions are displayed in Table 6. Firstly, from the energy utilization efficiency effect, as depicted by Columns (1) and (2), the coefficients for EPT (Tax_did), at the 1% level, remain notably positive. In a superior way, it reveals that EPT greatly promotes energy utilization efficiency, which is conducive to reducing energy utilization costs, promoting economic growth, and simultaneously decreasing environmental issues arising from energy usage. Most vitally, as noted in Column (3), from the findings for EPT on carbon emissions at the city level via partitioning low and high groups defined by the 30% and 70% quantiles of energy utilization efficiency, the coefficient for EPT on carbon emissions, at the 1% level, remains −0.201 in the high group for energy utilization efficiency (Tax_did_high), but is positive in the low group (Tax_did_low). That is to say, it indicates that the higher the scale upon energy utilization efficacy at an area scale is, the greater the benefit of EPT in reducing carbon emissions is, thus verifying research Hypothesis 2 (H2). In addition, the Wald statistic, as claimed by Column (3), denies the linear constraint null hypothesis for EPT, notably revealing that the method grouped by energy exhibits rationality.

Secondly, from the green technological innovation, with the results presented in Columns (4) and (5), the coefficients for EPT, at the 1% level, stay fairly positive. In a finer way, it reflects that EPT greatly promotes green technology innovation, fully lessening carbon emissions and improving environment ecology. Furthermore, as depicted in Column (6), from the regression results for EPT via splitting into two groups attested by the 30% and 70% quantiles of patent, the coefficient for EPT, at the 1% level, exhibits −0.086 in the high group for green technology innovation (Tax_did_high), and not significant in the low group (Tax_did_low). As it were, with the boost for green technology innovation, EPT can notably suppress carbon emissions and improve environment quality, thus verifying research Hypothesis 3 (H3). Moreover, the Wald statistic listed in Column (6) rules out the linear null hypothesis for EPT, greatly disclosing that the model settings grouped by Patent remains reliable as such.

Thirdly, from the industrial structure upgrading effect, being presented by Columns (7) to (8), the coefficients for EPT on industrial upgrading, at the 1% level, are distinctly positive. As plainly evident, it claims that EPT fairly promotes industrial upgrading, in a finer way, reducing carbon emissions alongside improving ecological situation in the area. Furthermore, based on the findings for EPT on carbon emissions grouped by a rate upon industrial upgrading, being exhibited in Column (9), from a 1% statistic perspective, the outcome for EPT serves −0.089 in the high industrial structure group grouped by Ind_up (Tax_did_high), yet exhibits not significant in the low group (Tax_did_low). That being the case, it reveals that, faced with the service-oriented growth and a more advanced industrial structure, EPT to a greater extent makes it easier to lessen carbon emissions at the city level, thus proving Hypothesis 4 (H4) in the process. Likewise, the Wald statistic listed in Column (9) to some extent rejects the linear constraint null hypothesis for EPT, subtly meaning that this practice grouped by Ind_up stays reasonable.

5.7. Spatial Spillover Effects Analysis

As is well known, carbon dioxide will undergo spatial flow with production activities such as factor flow and industrial transfer and will also diffuse among areas with changes in climate conditions. More importantly, the successful experience of promoting local carbon reduction through the set up of pilot cities may stimulate imitation and learning motivation in neighboring areas, leading to spatial spillover effects from pilot policy. Therefore, to disclose whether EPT can create a radiative benefit of carbon emissions reduction in neighboring regions, pertaining to Guo et al. [64], based on the adjacency relation alongside geographical distance between areas, two spatial weight matrices are constructed for spatial econometric analysis, involving spatial contiguity weight matrix (W1) and geographical distance weight matrix (W2). As depicted in Figure 7 and

Table 7. Spatial regression results for EPT on urban carbon emissions at the city level.

Variable	Spatial Contiguity Weight Matrix (W1)				Geographical Distance Weight Matrix (W2)
	Coefficient	Direct	Indirect	Total	Coefficient	Direct	Indirect	Total
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Tax_did	−0.092 ***	−0.093 ***	−0.035 ***	−0.128 ***	−0.097 ***	−0.099 ***	−0.595 ***	−0.694 ***
	(−7.08)	(−6.87)	(−6.24)	(−6.83)	(−7.46)	(−7.23)	(−3.27)	(−3.68)
ρ	0.289 ***				0.854 ***
	(17.17)				(26.08)
Control	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
ID FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.218				0.320
Log-Lik	917.841				887.261
Obs	4794	4794	4794	4794	4794	4794	4794	4794

Note: t value included in parentheses. In a finer way, at the 1% level, significance is reflected by ***.

, the empirical results for global Moran testing and spatial DID model are analyzed.

As noted in Figure 7, it reveals the global spatial autocorrelation testing outcomes for PCE, where the shaded areas reflect the 95% confidence interval for Moran’s I. To be more specific, it can be seen that whether using matrix W1 or W2, the values for global Moran’s I testing, from a statistical view, retain above zero at the 5% level, meaning that urban carbon emissions exhibit spatial dependence. That is to say, a rise in local carbon emissions notably yields a rise in neighboring areas. Thus, it is vital to employ spatial DID technique to further argue the benefit of EPT execution upon carbon emissions reduction.

As stated in Table 7, Columns (1)–(8) disclose the results of spatial spillover effect testing for EPT upon urban carbon emissions. To be more precise, from the estimated coefficients of EPT (Tax_did), whether using matrix W1 or W2, the impacts for EPT on carbon emissions remain notably negative from a statistical view, revealing that EPT to some extent reduces urban carbon emissions. Meanwhile, the spatial autoregressive coefficients (ρ) are greatly favorable, further proving the positive spatial dependence of carbon emissions as disclosed in Figure 6. Moreover, by using the partial differential decomposition technique, we further split the carbon reduction benefits for EPT into total effect (Total), indirect effect (Indirect), and direct effect (Direct). It can be seen that whether using matrix W1 or W2, the coefficients of Direct, Indirect, coupled with Total for EPT are greatly opposite, thus confirming the Hypothesis 5 (H5). This means that after considering the policy spillover effects, EPT not only subtly lessens carbon emissions at the city level, but also promotes imitation and learning in neighboring regions, such as driving innovation in green technology and upgrading industrial structure, forming a radiation impact upon carbon reduction in surrounding areas, and ultimately creating a driving benefit of carbon reduction in all areas.

6. Discussion

Taxation is the foundation and important pillar of national governance, and EPT, as an important green tax system, is more flexible than administrative command-based environmental regulation. It plays a crucial role in attaining low-carbon transition and high-quality development at a city scale. This study offers robust evidence that EPT can greatly lessen urban carbon emissions, thus confirming Hypothesis 1, which is also similar to the findings from Collins et al. [35]. The reason is that EPT is different from administrative command-based environmental regulation. It does not force firms to stop or reduce production but offers firms with the opportunity to choose production strategies through a price mechanism. To reduce pollutant emissions and pay less environmental taxes, firms need to adjust their production based on the balance between costs and profits. Therefore, although EPT execution increases the burden on firms, it does not actually suppress their production. Instead, it stimulates their technological innovation, energy conservation and emission reduction, and green production, thereby improving production efficiency and promoting the long-term reduction of carbon emissions. The findings disclose that it is of great significance to optimize the green tax system and explore the use of market mechanisms to reduce carbon emissions in the context of the dual carbon target and high-quality economic development.

This study reveals that EPT policy affects carbon emissions at an urban scale through enhancing energy efficiency (Hypothesis 2), spurring green technological innovation (Hypothesis 3), and promoting industrial structure upgrading (Hypothesis 4), thus confirming the theoretical analysis. From energy efficiency, EPT imposes additional costs on energy usage, offers a driving force for firms to optimize production processes and invest in energy-saving technologies. Especially, EPT includes tax incentives for the adoption of renewable energy, motivating firms to develop and apply cleaner technologies to enhance energy utilization while reducing dependence on fossil fuels. Similar to existing research, EPT promotes economic activity towards more energy-efficient sectors and green industries, thereby reducing environmental pollution [19]. From green technological innovation, EPT can increase the cost of carbon emissions, prompting firms to enhance their profitability and competitiveness by expanding R&D investments for energy-saving and emission reduction. The use of low-carbon technology not only helps firms meet carbon emission rules but also brings long-term economic benefits to firms by lowering operating costs, further enhancing their focus on green technologies. Similar to existing research, green tax policies can effectively stimulate green technology innovation, thereby reducing pollutant emissions [52]. From industrial structure upgrading, EPT stimulates firms to expand R&D funds in emission reduction and low-carbon technologies, use advanced pollution control technologies and equipment, achieve more efficient and cleaner production methods, and promote the evolution of labor-based industrial structures to technology-based industries, thus reducing carbon emissions. Therefore, by disclosing the impacts through three channels, we offer a theoretical basis for improving EPT system and fully leveraging the carbon reduction effect of EPT.

This study reveals the heterogeneous impacts of EPT on urban carbon emissions, and the effectiveness of policies varies depending on geographical location, innovation level, resource endowment, and industrial structure. From geographical location, EPT exhibits a significant carbon reduction effect in central and western cities, while the inhibitory effect in eastern cities is not significant. The insignificant carbon reduction effect for EPT in eastern cities is attributed to the formation of a relatively mature green industrial structure and the transfer of a large number of polluting industries to other regions. From the level of innovation, EPT exhibits a significant carbon reduction effect in innovative cities, but the impact in non-innovative cities is poor. A plausible claim for non-innovative cities is attributed to the lack of vital technological infrastructure and green innovation capabilities. From resource endowment, EPT exhibits a significant carbon reduction effect in non-resource-based cities but is not significant in resource-based cities. The main reason lies in the fact that resource-based cities may easily form a growth paradigm depicted by high energy usage and high pollution discharges, resulting in limited emission reduction effects of EPT. From the industrial attribute, EPT exhibits a significant carbon reduction effect in non-old industrial base cities but is not significant in old industrial base cities. The insignificant effect of EPT in old industrial base cities is attributed to outdated and high-emission industrial systems, making it difficult to implement green innovation and industrial low-carbon transition. The above findings are of great significance for a deeper understanding of the nexus between environmental regulation and carbon emissions [63].

This study reveals that EPT has exhibited significant spatial spillover effects; that is, the execution of local EPT policy can greatly affect carbon reduction in spatially correlated regions, thus confirming Hypothesis 5. Theoretically, after considering policy spillover effects, EPT not only subtly lessens local carbon emissions at the city level, but also promotes imitation and learning motivation in neighboring areas, such as promoting green technology innovation and industrial upgrading while enhancing government spending on environmental protection projects, thus forming a radiation impact on carbon reduction in adjacent areas. Although the existing research discerns that the externalities of environmental pollution imply that spatial dependence attributes must be considered when studying carbon emissions, omitting spatial correlation and spillover effects can easily lead to estimation bias for policy effects, which is unable to assess the carbon reduction effects for EPT and define carbon responsibilities; spatial DID models are still rarely adopted to explore the carbon reduction effects for EPT in practice. This study discloses the spatial spillover effects for EPT upon carbon emissions at the city level, which to some extent helps to identify the carbon reduction effects for EPT and offer empirical evidence for improving green tax policies system to realize the dual carbon target.

Although the study has comprehensively examined how the EPT scheme reduces carbon emissions at the prefecture level, it must be acknowledged that the empirical procedure still has two limitations. First, we failed to consider the impact of policy linkage on carbon emissions at the city level, which can be addressed through introducing other related policies such as low-carbon city pilot and carbon emission trading pilot to disclose the carbon reduction effect of environmental policy synergy in the future. Second, the dataset used in the empirical tests does not involve micro-level fields, and subsequent research will consider micro-survey data or detailed county-level statistics to enhance the robustness and comprehensiveness of the empirical outcomes.

7. Conclusions and Policy Recommendations

The relentless surge in carbon emissions, especially for the developing countries, has steadily fueled critical issues like global climate change and environmental degradation. To tackle this dilemma, the Chinese government is actively pursuing various environmental protection measures including the EPT scheme to direct economic development towards a greener, low-carbon path. Against this backdrop, using the 2006–2022 dataset from 282 Chinese cities at the prefecture level, this study adopts an I-DID technique to delve into how EPT affects urban carbon emissions. The findings reveal that EPT execution to some extent can substantially lessen carbon emissions at the city level, which remain robust through parallel trend testing, excluding other related policies, placebo testing, etc. Looking at heterogeneity, the benefit of the EPT scheme on carbon emissions varies considerably, due to the stark differences in innovation capacity, resource availability, and the level of industrial development across different cities. Especially in innovative cities, central and western cities, non-industrial-based cities, alongside non-resource-dependent cities, EPT packs a bigger punch when it comes to curbing urban carbon emissions. From the impact mechanisms, EPT has greatly reduced urban carbon emissions by promoting industrial structure, enhancing energy usage efficacy, as well as advancing green technology innovation. More importantly, from the spatial spillover effects, EPT not only subtly lessens local carbon emissions at the city level, but also promotes imitation and learning in neighboring regions, such as driving green innovation as well as preserving industrial framework, forming a radiation impact upon carbon reduction in surrounding areas, and ultimately creating a driving benefit of carbon reduction in all areas.

To fully reconcile economic growth with pollution control, deepening the reform for green tax policy system is of great significance for improving the efficacy of EPT on carbon emissions. So, with the findings mentioned above, here are a few policy plans.

Firstly, enhance the green tax system with EPT as the core, and accelerate the green and low-carbon transition of the tax system. As a key green tax system, the operation effect of EPT largely depends on the tax rate, enforcement intensity, and coverage of taxable pollutants. To fully realize the carbon reduction effect of EPT, first, governments should gradually expand the scope of EPT collection. For example, such pollutions from volatile organic compounds, radioactive substances, pesticides, forests, and grasslands should be included in EPT collection. Meanwhile, many tax items such as resource tax, ecological tax, and carbon tax should be added to EPT system to expand the coverage of collection and help achieve carbon reduction in multiple dimensions. Second, governments should fully consider the local environmental carrying capacity, pollutant emission status, and control needs, and timely raise the rate of EPT, with a focus on increasing the illegal costs of firms, especially high polluting and high emission firms. By raising the tax rate, firms can be incentivized to enhance carbon emissions efficiency, thus achieving the carbon control effect of green taxation. In addition, it is essential to enhance the cooperation between the environmental protection bureau and the tax bureau; establish a big data comprehensive management platform, environmental data center, and intelligent carbon emission supervision platform; increase the investigation and punishment of carbon emission violations; enhance the efficacy and deterrence of EPT; and encourage firms to carry out green transition as well as technology innovation breakthroughs.

Secondly, fully assess regional heterogeneity, combine resource endowment and urban characteristics, and implement a diversified and differentiated EPT system according to local conditions. Since cities vary widely in terms of location, innovation capability, resource endowment, and industrial mode, local governments should adopt suitable carbon reduction plans tailored to its situation and find a new path for low-carbon development. In the central and western regions and non-innovative cities, governments can properly boost the basic tax rate of EPT and reduce the pollution control costs of firms through the form of tiered reduction and exemption of EPT, guide firms to increase green investment, and encourage talent and innovative resources to flow to the central and western regions. For the eastern region and innovative cities, in addition to providing reasonable tax incentives to firms that reduce emissions significantly, governments can support the creation of green innovation demonstration firms, offer subsidies or policy incentives, and encourage outstanding firms to raise green innovation investment, thus promoting the formation of green industrial system. In resource-based cities, governments should properly raise the standards of EPT, levy tax items such as resource extraction tax, groundwater tax, forest tax, and reduce the ratio of high polluting and high energy consuming industries, thus attaining the low-carbon transition of industrial structure. Moreover, to break the path dependence, it should provide financial support for resource-based cities, guiding resource factors to cluster in service and high-tech industries through infrastructure investment, and weakening the dependence on fossil fuels, thus reducing carbon emissions. In old industrial base cities, governments should implement differentiated and incentive measures based on the needs of industrial upgrading. While strictly imposing EPT, firms that meet carbon emissions standards should be appropriately exempted from EPT. Meanwhile, it should increase tax incentives and financial support and use policies such as income reduction and exemption for energy-saving and environmental protection projects, value-added tax refunds, etc., to guide firms to transform towards green development.

Thirdly, to forcibly achieve decarbonization, a synergistic environmental control system integrating energy transition, green technology innovation, and industrial upgrading must be formatted. Specifically, governments should optimize energy supply structure, explore the exit path of coal-fired power, establish an energy system dominated by clean energy such as wind, hydropower, and solar power; advocate green living, low-carbon consumption, low-carbon travel; and formulate policies including energy efficiency standards and energy audits to promote energy market reform. Concurrently, governments should offer fiscal support for firms engaged in energy conservation, and clean production, and use tax incentives, subsidies to support firms, universities, and R&D institutions in building green innovation cooperation platforms. Moreover, it should accelerate the formation of institutional mechanisms that encourage the promotion and application of green innovation, cultivate talents in green technology innovation such as carbon capture and storage, promote the free flow of green innovation factors, and thus improve urban green technology innovation. In addition, governments should coordinate EPT policy and other industrial policies to guide urban industries towards low-carbon development. To be specific, it should accelerate the low-carbon transition in traditional industrial fields such as steel smelting and oil extraction, strictly implement equal or reduced capacity replacement for high energy consuming and high emission projects. Meanwhile, it should vigorously develop green and low-carbon industries, accelerate the strategic emerging industries such as new energy and biotechnology, and promote the deep integration of green and low-carbon industries with emerging technologies, thereby promoting the industrial upgrading of the whole society.

Finally, enhance communication and cooperation among cities, and establish a cross-regional synergic system guided by carbon control. Governments should abandon the fragmented and beggar-thy-neighbor territorial management model and create a decentralized cross-regional joint prevention and cooperation model, establish a normalized regional feedback mode and information sharing platform, and promote cross-regional joint control of carbon emissions. To be specific, to avoid the nearby transfer of high polluting industries in pilot cities and the negative impact of policy siphoning on low-carbon development in adjacent cities, non-pilot cities should raise the entry threshold for polluting firms, prevent the large introduction of polluting firms, and lessen the negative impact of pollution industry transfer on carbon emissions control from a supervision view. Meanwhile, governments should build green technology innovation platforms, enhancing communication and cooperation between pilot cities and neighboring cities, promoting the cross-regional flow of innovation factors, realizing the sharing of innovation products among cities, filling the gaps in carbon emission control in surrounding cities, and leveraging the radiation effect of pilot cities on neighboring cities in green technology innovation and pollution control experience. In addition, governments should propel the regional coordinated development, establish a carbon emission verification and coordinated control system, especially a pollution monitoring network, form a cross-regional joint prevention and control system, promote the layout of low-carbon industries and regional environmental joint control, steadily optimize and integrate regional innovation resources, achieve complementary advantages and industrial structure upgrading, and promote the comprehensive green transformation of the economy and society.