Market-Based Environmental Regulation and Green Transformation of Manufacturing Enterprises: Evidence from China’s SO₂ Emission Trading

Abstract

In the context of China’s sustainable development goals, this study investigates how market-based environmental regulation, specifically SO₂ emission trading, drives green transformation and upgrading in manufacturing firms. We argue that emission trading can spur firms to optimize their service portfolios and improve their embedded services, thus facilitating a shift toward greener operations—while its effect on a firm’s technological innovation capacity remains ambiguous. Using panel data from Chinese manufacturing firms, our empirical findings show that emission trading significantly promotes service transformation through optimizing embedded services and reducing mixed, noncore services but does not necessarily encourage increased R&D investment or breakthrough innovation. This suggests many enterprises prefer end-of-pipe pollution control measures rather than structural technological upgrading under regulatory pressure. Our results provide insights into refining emission trading mechanisms and designing supportive policies to better channel firm-level transformation toward sustainability.

1. Introduction

The development of the manufacturing industry has made a significant contribution to China’s industrialization and modernization process, but the extensive growth model with high pollution and high energy consumption has also exacerbated environmental pollution, greatly increased the pressure for transformation and upgrading of the manufacturing industry, and seriously threatened the sustainable development of the economy [1]. The “Fourteenth Five-Year Plan” of the Party clearly proposes to “win the battle against pollution prevention and control, promote the timely compliance of industrial pollution sources with emission standards, and promote the marketization of emission rights, energy rights, water rights, and carbon emission rights.” Under the development concept of “Green mountains and clear waters are as good as mountains of gold and silver,” establishing a sound ecological environment governance system and giving play to the decisive role of the market mechanism in resource allocation is the key to promoting a “win-win” situation between the transformation and upgrading of the manufacturing industry and environmental friendliness and to achieving high-quality economic development and high-level environmental protection [2]. Clarifying the relationship between pollution control and the transformation and upgrading of the manufacturing industry, and studying the impact of environmental regulation policies on the transformation and upgrading, are of great practical significance in supporting the high-quality development of the manufacturing industry and promoting the transformation and upgrading of enterprises.

Reviewing the development history of China’s environmental governance system, command-and-control regulation policies were predominant in the early stage, supplemented by the pollutant discharge fee system [3]. As environmental pollution problems have become increasingly serious, market-based environmental regulation measures have gradually played a role, with the emission trading system being the most typical. The basic idea of the emission trading system is to establish a legal market for trading pollution emission rights on the basis of clearly defining the property rights of emission rights so as to minimize the total cost of pollution control for the whole society, thereby achieving the optimal allocation of environmental resources and the goal of controlling the total amount of pollution emissions [3]. In theory, the direct effect of the implementation of the emission trading system is to reduce the pollution emissions of enterprises and drive enterprises to adopt green production, while the indirect effect is to promote green innovation and industrial structure upgrading through resource flow, which brings welfare to the green development of society [4]. Then, has the emission trading system had the expected effect of controlling sulfur dioxide emissions? Accordingly, what impact has the emission trading system had on the transformation of micro-enterprises? Do micro-enterprises have the strategic choice of “proactive change,” either through green innovation to renovate outdated production equipment and processes or by shifting to service-oriented manufacturing and extending downstream in the industrial chain? This paper will discuss these issues in detail from both theoretical and empirical perspectives.

The relevant literature on the impact of environmental regulatory policies on industrial upgrading mainly revolves around two hypotheses. The first is the research based on the “pollution haven hypothesis” [5], which suggests that environmental regulations increase the production costs of manufacturing enterprises. To compensate for their efficiency and profit losses, enterprises make decisions on entry, exit, and industrial relocation through adjustments in production scale and optimization of resource allocation [5], thereby affecting the regional industrial structure [6,7]. The second is the research based on the “Porter hypothesis”, which examines the incentive effect of environmental regulations on green innovation and verifies their retroactive mechanism on industrial structure adjustment [8]. However, the empirical conclusions are not entirely unified. Some scholars find that environmental regulations have not achieved the “Porter effect” due to low market efficiency and relatively weak regulatory intensity [3], while other scholars have verified that in the long run, environmental regulations can achieve a win-win situation between environmental friendliness and economic development through green innovation [9].

With the pilot implementation of China’s emission-trading system, research has increasingly examined its environmental performance; Zhang et al. (2020) provide evidence that the trading policy contributed to emission reductions and improved market efficiency [10]. In terms of environmental performance, the existing literature mainly focuses on the effectiveness of market-based environmental regulatory instruments in reducing enterprise pollution, resulting in completely different research conclusions. Some scholars have affirmed the policy effects of emission trading, with evidence at the macro level showing that it can not only reduce the emissions of the pollutants traded [11,12] but also have a synergistic emission reduction effect on other air pollutants [13] and improve the energy consumption per unit of output in the pilot regions [14]. Discussions at the micro level are relatively few, but some research has shown that when market-based environmental regulations are combined with command-and-control environmental regulations, the sulfur dioxide emission reduction effects of enterprises are significantly enhanced [3].

As an important means of market-based environmental regulation, the effectiveness of the emission trading system not only lies in the short-term goal of reducing pollution emissions but also aims to achieve the long-term goal of win-win sustainable environmental improvement and green economic development. Existing research literature on economic performance is relatively scarce, mainly focusing on the indirect impact of the emission trading system on the development of green innovation in enterprises [15], labor employment demand [16], and foreign direct investment [17]. The relevant domestic and foreign literature on whether market-based environmental regulation can affect the development of green innovation in enterprises is relatively rich, but a consistent view has not yet been formed: some scholars believe that while emission trading has to some extent alleviated the inefficiency of command-and-control regulation, its role in improving the level of green technology in the pilot regions is minimal, and the Porter effect has not yet formed in China [18]; in contrast, some scholars support the validity of the “Porter hypothesis” in China, proposing that market-based environmental regulation can significantly improve the efficiency of green innovation in industries, and help enterprises improve their operational performance through the innovation effect [19].

Overview of the existing literature: Although the research on the environmental governance effects of the emission trading system is relatively rich, the exploration of the transformation and upgrading paths of manufacturing enterprises is insufficient, and the sporadic literature only involves the analysis of enterprise innovation performance and total factor productivity [16,18], without delving into the impact of the emission trading system on the strategic choices of manufacturing enterprises. Whether the emission trading system can bring the expected pollution reduction effect to enterprises is certainly a topic that needs to be discussed, and scholars have already paid sufficient attention to it. However, faced with market-based incentive environmental regulation measures, what kind of transformation and upgrading path will micro-enterprises choose? This is a question with more academic value and policy implications, and the existing literature has not given a clear answer. Addressing the above questions faces two challenges: (1) Break through the discussion of the role of environmental regulation policies in industrial transformation and upgrading, focusing on the mechanism of market-based environmental regulation affecting the transformation and upgrading of micro-enterprises, and theoretically explore possible path choices; (2) Expand empirical research at the regional level- or industry level-data, overcome the limitation of being unable to distinguish whether the change in environmental quality comes from government environmental governance measures or micro-individual governance behavior, and accurately identify the coping strategies of manufacturing enterprises [20,21].

Overcoming the aforementioned two major challenges, this paper takes the emissions trading system as the object of analysis and discusses the path selection of manufacturing enterprises’ transformation and upgrading under market-oriented environmental regulation from both theoretical and empirical perspectives. This study finds: First, the emissions trading system has significantly reduced the emission levels of sulfur dioxide by manufacturing enterprises and has produced a synergistic emission reduction effect, with the results remaining robust after addressing endogeneity issues. Second, the emissions trading has increased the service business upgrade oriented by the extension of the manufacturing enterprise value chain, without significantly affecting the green innovation transformation, indicating that under market-oriented environmental regulation, micro-entities will prioritize service transformation towards the downstream of the value chain, rather than innovation transformation towards the upstream of the value chain. Third, the emissions trading promotes service optimization by increasing customer concentration and fixed asset investment, reducing the proportion of current assets and marketing intensity. The level of embedded service integration of upstream enterprises in the production chain is significantly enhanced, and the digitalization level has strengthened the service optimization and upgrading effect of enterprises. Fourth, in the face of market-oriented environmental regulation, manufacturing enterprises have not increased R&D investment but have relied on end-of-pipe pollution control to achieve emission reduction targets. A good technical foundation will allow enterprises to “face difficulties head-on,” while market power hinders the technological innovation and upgrading of enterprises through cost transfer.

Compared to existing literature, the potential marginal contributions of this paper are as follows. In theoretical analysis, existing literature lacks discussion on the issues related to the transformation and upgrading of manufacturing enterprises [14,18]. This paper reveals the mechanism of path selection for the transformation and upgrading of manufacturing enterprises under market-oriented environmental regulation from the perspective of the value chain. It innovatively discusses the service optimization effect extending downstream in the value chain and the innovation incentive effect climbing upstream in the value chain, proposing the research hypothesis that manufacturing enterprises will prefer service transformation and conditionally choose technological innovation and upgrading. In terms of identification methods, based on whether it is conducive to the vertical extension of the value chain, all service businesses involved in manufacturing enterprises are classified into embedded services that extend the value chain and mixed services that are unrelated to the value chain. This compensates for the insufficiency of traditional manufacturing service integration measurement methods based on input-output tables, which can only measure the level of input serviceization and cannot identify the degree of output serviceization, thus discussing the extension of the manufacturing enterprise value chain. It expands the research methods in the field of integration between manufacturing and service industries [22]. In empirical and applied research, the Chinese enterprise pollution database is matched with microdata of listed manufacturing companies, comprehensively and meticulously examining the impact of emissions trading on pollution emissions and path selection of manufacturing enterprises’ transformation and upgrading. It accurately assesses the mechanism of action, providing micro evidence for the first time for empirical studies on market-oriented environmental regulation and industrial transformation and upgrading. It offers new policy insights for further improving environmental governance mechanisms in the new era of socialism with Chinese characteristics, and for the synergistic role of effective markets and proactive governments. In recent years, a growing body of empirical research has extended the investigation of market-based environmental regulation beyond emission reduction to its potential role in driving firm-level transformative change and innovation. For instance, Luo & Wang (2023) document that stricter environmental regulation fosters green innovation and enhances high-quality enterprise development in Chinese listed firms [23]; Wang et al. (2024) show that environmental regulation enhances the green economy partly through corporate green technological innovation [24]; Su (2025) finds that the implementation of China’s total energy consumption targets significantly promotes firm-level green innovation across pollution-intensive industries [25]; Chen et al. (2025) reveal that voluntary regulatory instruments such as green factory certification positively influence both the quantity and quality of green technological innovation [26]; Wang & Xing (2025) provide micro-evidence from manufacturing enterprises that China’s carbon emissions trading pilot policy stimulates green technological innovation [27]. Collectively, these studies highlight an evolving research frontier that links market-oriented environmental regulation with firm transformation, technology upgrading, and value-chain repositioning. However, the extant literature still lacks a focused examination of how manufacturing firms strategically respond to emission trading through upgrading their service-oriented business models or integrating downstream manufacturing services, especially in the context of Chinese SO₂ emissions trading. Addressing this gap, the present study examines the roles of service transformation and embedded service integration in manufacturing firms under SO₂ emissions trading, thereby offering novel micro-level evidence for the ‘win-win’ paradigm of environmental protection and industrial upgrading.

2. Theoretical Analysis and Research Hypotheses

2.1. Path Selection Mechanism

Command-and-control environmental regulations have a strict punishment mechanism, but they provide inadequate incentives for enterprises to undergo strategic transformation, and it is difficult for them to generate effective compensation for innovation [2]. Emission rights trading systems (such as China’s pilot carbon-emissions trading policy) create market-based incentives for firms, enabling them to reduce emissions or innovate in response to cost-/revenue signals [28]. After the implementation of the “emission rights trading” system, regardless of the choice made by enterprises, they can minimize their own pollution control costs in the short term. From a long-term perspective, as a tradable resource that reflects the scarcity of environmental capacity, the price of emission rights will continually rise, and the costs of pollution reduction and environmental management for enterprises will also continually increase, leading to the elimination and exit from the market of inefficient polluting enterprises [29]. The implementation of the emission rights trading system is essentially cost-effective, as it solves the problem of inefficient allocation of emission rights, and can effectively stimulate the energy-saving and emission reduction potential of enterprises, as well as better motivate enterprises to make strategic adjustments to achieve cleaner production [30,31].

The basic principles for selecting the optimal path for manufacturing industry transformation and upgrading are the improvement of resource allocation efficiency and the enhancement of production efficiency [32]. Manufacturing enterprises will engage in high-value-added production and business activities by enhancing their market capabilities and technological capabilities, thereby improving their competitiveness to achieve transformation and upgrading [33]. From a value-chain perspective, service integration enables the extension of the industrial chain and functional upgrading toward downstream value-chain segments [34], while green innovation is conducive to deepening the degree of participation in the industrial chain and completing product upgrades towards the upstream value chain [35]. Accordingly, the transformation and upgrading of the manufacturing industry should include two choices: (1) through low-risk and low-investment service integration, to explore new markets, increase customer satisfaction, and promote the deep integration of manufacturing and modern service industries [36]; (2) through high-investment and high-risk green technology innovation, to improve product quality, resolve the overcapacity issue in traditional manufacturing industries, upgrade to higher processing and higher value-added, and ultimately achieve excess profits [2]. Whether it is the service integration to realize functional value, or the green innovation to pursue product value, both can enhance the core capabilities of enterprises, create customer value and competitive advantages (see Figure 1), improve the problem of lock-in at the low-end of the value chain, and promote the rise in the global value chain division of labor position of manufacturing enterprises.

Under the influence of the emission trading system, external environmental pressure will increase the opportunity cost of the manufacturing process for micro-enterprises, which will drive manufacturing enterprises to enhance their autonomous emission reduction motivation [37]. Spurred to upgrade: a review of triggers and consequences of industrial upgrading in the global value chain literature [37]. Once the inherent incentive effect of market-based environmental regulation is fully realized, the benefits of manufacturing transformation and upgrading will far outweigh the increase in their pollution control costs [38]. Considering the cost constraints and the drive to maximize profits, manufacturing enterprises will fully utilize their own resource advantages and operational capabilities on the basis of emission reduction and make proactive strategic adjustments [39], either by choosing to extend their value chain downstream through service transformation or to climb upstream through innovation upgrading to offset the cost increase brought about by the emission trading system and obtain the potential benefits of enterprise value chain extension, thereby realizing transformation and upgrading [35,40]. Therefore, Hypothesis 1 is proposed.

Hypothesis 1:

Under the emissions trading system of market-oriented environmental regulation, manufacturing enterprises will choose to extend downstream in the value chain through service transformation, or to climb upstream in the value chain through innovation and upgrading.

2.2. Service Optimization Effect

The implementation of the emissions trading system imposes operational pressure on manufacturing firms, raising their awareness of emission reduction and triggering responses in production processes and business models; empirical evidence indicates that pilot schemes in China have prompted firms to increase low-carbon innovation activities rather than simply continuing high-pollution operations [41]. While servitization (service transformation) remains a longer-term strategic option, in the short term, firms appear to respond by reallocating their innovation resources and shifting toward less-carbon-intensive production. As the cost increase brought by emissions trading to enterprises, it increases the risk and difficulty of technological innovation for manufacturing enterprises. It is difficult to form effective green technological innovation products in the short term. Focusing on meeting the service needs of downstream customers and consumers in the industry chain has become an effective response. As the cost increase brought by emissions trading to enterprises, it increases the risk and difficulty of technological innovation for manufacturing enterprises. It is difficult to form effective green technological innovation products in the short term. Focusing on meeting the service needs of downstream customers and consumers in the industry chain has become an effective response. The transfer of the manufacturing enterprise value chain to service business and the strengthening of the participation of downstream customers form a close connection between the manufacturing business and customer service. This helps manufacturing enterprises to better meet customer needs and enhance their ability to cope with external market risks [42].

Due to the existence of resource constraints, when manufacturing enterprises face conflicts in the introduction and adjustment of service business, conflicts must be resolved through the principle of strategic consistency [43]. To ensure the strategic goal of value creation in the face of the implementation of the emissions trading system, manufacturing enterprises must not only make strategic decisions on whether to transform to services but also optimize the allocation of limited resources and choose the type of business in the field of service expansion. According to the standard of contribution to the value enhancement of manufacturing enterprises and referring to the concept of service relevance [44], the service business involved by manufacturing enterprises can be divided into two types: one is the embedded service that revolves around the core manufacturing product and is conducive to the extension of the value chain. By integrating manufacturing and service resources, it enhances the added value of products and customer satisfaction, thereby strengthening the competitive advantage of manufacturing enterprises; the other is the embedded servitization that deep-embeds services into the manufacturing value chain, whose original intention is to reallocate enterprise resources through unrelated diversification to obtain more profits [45]. The implementation of the emissions trading system will have heterogeneous impacts on the hybrid servitization, with weak strategic linkage to the core manufacturing business and the mixed services that are unrelated to the value chain.

The creation effect of embedded services. Under increasing regulatory and cost pressures, manufacturing firms are turning to services: manufacturing firms under external regulatory and cost pressures increasingly integrate services with their core products to enhance added value and customer satisfaction [46]. To compensate for the increased control costs and losses due to reduced production and to improve the added value of production and operational performance [47], enterprises will focus on customer needs, optimize the allocation of limited resources, and expand into cleaner embedded service segments centered around core products. On the one hand, establishing a more stable relationship with customers, obtaining more product demand information from customers, and enhancing customer satisfaction and product added value; on the other hand, increasing customer loyalty and market demand stickiness, promoting the enhancement of market monopoly power, and seeking greater profit margins. After the implementation of the emissions trading system, manufacturing enterprises can reduce the demand price elasticity of products through the integration and transformation of embedded services, thereby increasing the enterprise’s cost transferability and passing on the additional regulatory costs caused by the emissions trading to downstream consumers. Considering the potential benefits of embedded service integration and the relative improvement in cost transferability, the motivation for managers to undergo embedded service transformation is also enhanced [48].

The Crowding-Out Effect of Mixed Services. According to the resource-based view, the value creation capability will only be enhanced when the strategic efforts of enterprise service integration are consistent with its environmental conditions [49]. Although mixed services can bring certain profit growth to enterprises in the short term, service businesses unrelated to the core products are not conducive to the extension of the manufacturing value chain and may disperse limited enterprise resources. In the long run, on the one hand, the inability of enterprises to meet service demands related to their products will lead to a decrease in consumer satisfaction and loyalty, thereby reducing product competitiveness and enterprise performance. On the other hand, the provision of mixed services can lead to the loss of strategic focus for enterprises, causing them to lose core competitiveness and have a negative impact on the value creation of their main business [50]. Enterprises can optimize resource allocation and reduce operating costs by separating service elements from mixed services [51], enhancing the division of labor in the enterprise’s value chain, and thereby increasing enterprise resilience. After the implementation of the emissions trading system, both pollution control and the purchase of emissions rights will affect the operating cash flow, manifested as a decrease in the proportion of current assets of enterprises. Faced with increasingly scarce disposable resources, manufacturing enterprises are motivated to optimize their business structure, divest some mixed services that lack competitive advantage and have higher operational risks, and allocate limited resources to embedded services related to core products, thereby achieving the extension of the manufacturing enterprise’s value chain. Therefore, Hypothesis 2 is proposed [36].

Hypothesis 2:

The “emissions trading” system can bring about a service optimization effect in manufacturing enterprises, creating embedded services that are conducive to the extension of the value chain and divesting mixed services that are unrelated to the value chain.

2.3. Innovation Incentive Effect

To explore the impact of “emissions trading” on the transformation and upgrading of the manufacturing industry, it is necessary to examine not only the service capabilities of manufacturing enterprises in addressing customer needs but also the choices regarding the enhancement of core technological innovation capabilities in the production process of micro-enterprises. The “Porter Hypothesis” theory suggests that the emissions trading system will encourage enterprises to engage in green technological innovation, thereby offsetting the additional pollution control costs imposed by environmental regulations and enhancing the competitive advantage of enterprises [3].

The emission trading system will increase the opportunity cost of the manufacturing process, enhance the potential benefits of climbing upstream to the innovation process, forcing enterprises to pursue green innovation. This allows enterprises to improve the competitiveness of their manufacturing products while ensuring pollution reduction, thereby offsetting the cost increase brought about by the emission trading system. Specifically, for the demand side of emission permits, enterprises need to purchase a certain amount of emission permits for their energy consumption and pollution emissions, which increases the manufacturing cost of the enterprise, also known as the “compliance cost”. For the supply side of emission permits, manufacturing enterprises need to increase their investment in emission reduction and pollution control, which will cause the labor capital and other factors originally used for manufacturing production to shift towards pollution control, thereby crowding out other profitable business investments to some extent, also bringing additional cost burdens to the enterprise, known as the “pollution control cost”. Enterprise innovation activities help to improve production efficiency and product quality, which are the key drivers for manufacturing enterprises to gain competitive advantages and market power [52]. Considering the increase in these two types of costs, and to safeguard the long-term economic interests of the enterprise, manufacturing enterprises have the motivation to reallocate production factors and resources and increase resource investment in high-efficiency and clean production and operation departments [13], thereby forming an innovation compensation effect.

The premise for the validity of the Porter hypothesis is that policy instruments can provide enterprises with significant economic incentives or cost pressures, and manufacturing enterprises need to use R&D investment or clean production to offset the cost increase brought about by environmental regulations [53]. After the implementation of the emission trading system, although manufacturing enterprises are motivated to adopt new production technologies to reduce pollution emissions [54], due to the conversion costs of green technology innovation and the high investment, high risk, and long payback period characteristics of product innovation activities, under the condition of maintaining the total output, manufacturing enterprises will first consider low-cost and rapid end-of-pipe pollution control. Strengthening end-of-pipe pollution control can quickly improve the enterprise’s pollution treatment capacity, which not only helps to reduce the compliance costs in the production and manufacturing process but also overcomes the high investment and uncertainty risks brought by product innovation. Furthermore, it can further reduce the manufacturing costs of enterprises through waste recycling and improve the economic performance of an enterprise’s production and operation. Based on this, even with the innovation incentives brought by emission trading, manufacturing enterprises will prioritize a more gradual technological improvement path [55] by purchasing advanced environmental protection equipment and increasing their investment in end-of-pipe treatment to improve their pollution control capabilities and efficiency in order to meet the clean production standards.

The emissions trading system can only exert real pressure on enterprises’ green technology innovation when manufacturing enterprises find it difficult to pass on the increased costs caused by sulfur dioxide emissions to other enterprises, and the end-of-pipe treatment reaches its limit in pollution reduction capacity [56]; otherwise, the effectiveness of environmental regulation policies will be affected, and the compensatory effect of innovation will be difficult to achieve [53]. On the one hand, the upfront investment in enterprises’ technological innovation activities is huge, often with a high degree of uncertainty and along a payback period. Especially under the condition of relatively weak technological capabilities and prior accumulation, the opportunity cost of upgrading products through upstream R&D is higher [57], which leads many manufacturing enterprises to be unable to overcome the technological threshold. That is to say, the emissions trading system can only have an innovation incentive effect when manufacturing enterprises have a certain technological foundation. On the other hand, firms with lower market power in competitive environments may face greater disruption from cost pressures, and thus the implementation of market-based environmental regulation appears more likely to stimulate their motivation towards green innovation. Conversely, firms with higher market power, when confronted with compliance costs under an emissions-trading system, are more inclined to engage in cost-shifting strategies rather than pursuing green innovation [58]. Based on the above analysis, Hypothesis 3 is proposed.

Hypothesis 3:

The innovation incentive effect of “emissions trading” is conditional; manufacturing enterprises will prioritize end-of-pipe treatment. The stronger the technological foundation, the more motivated they are to compete in the market, and only then will the innovation incentive effect become apparent.

3. Data and Model Design

3.1. Model Specification

Emissions trading, as a market-oriented environmental regulation policy implemented in batches in pilot areas, provides an excellent “quasi-natural experiment” for the use of the difference-in-differences method in this paper. This paper considers the emissions trading pilot, which has been officially implemented in 11 provinces since 2007, as a quasi-natural experiment and uses the difference-in-differences (DID) method to examine the environmental governance effects of emissions trading and its impact on the transformation and upgrading of manufacturing enterprises. The basic model is as follows:

Y_it = β₀ + β₁ (Treat_it × Post_it) + αControl_it + δ_t + γ_i + ε_it

(1)

In the model, Y_it is the dependent variable, representing the sulfur dioxide emissions, level of service integration, and level of green innovation for firm i in year t. Treat is the treatment group variable, Post is the post-treatment period variable; Control includes a set of control variables, such as firm size (Size), debt-to-asset ratio (Lev), capital-to-labor ratio (Capla), equity concentration (Cocen), investment-to-expense ratio (Inexr), and growth in profitability (Grow); δ_t represents the time fixed effect, γ_i represents the firm fixed effect, and ε_it denotes the idiosyncratic error term.

From the perspective of identification methods in quasi-natural experiments, the evaluation of the “emissions trading” pilot program may be subject to sample self-selection bias, meaning that areas with relatively severe pollution emissions are more likely to be prioritized by the central government for the “emissions trading” pilot reform. As a result, in the evaluation process of the natural experiment, there will be a challenge in effectively separating the effects of the pilot policy from the specific attributes of the experimental group, making it difficult to accurately assess the net effectiveness of the policy. The dependent variables in this paper are divided into sulfur dioxide emissions, representing the effect of pollution reduction, and service integration and green innovation, indicating the transformation and upgrading of the manufacturing industry. On the one hand, regarding the effectiveness test of pollution reduction, to address the potential sample self-selection bias, the air circulation coefficient is chosen as the instrumental variable for inclusion in the sulfur dioxide emissions trading pilot in the baseline regression, resolving the endogeneity issue of the experimental group’s pilot area selection, and thereby obtaining an accurate policy assessment effect. On the other hand, the path of transformation and upgrading of the manufacturing industry is not closely related to the natural experiment itself (“emissions trading” policy) and has a strong exogeneity. Therefore, by examining the impact of market-oriented environmental regulation on the transformation and upgrading of the manufacturing industry through the “emissions trading” pilot in 11 provinces, the assumptions of a “quasi-natural experiment” are met, allowing for a more accurate assessment of the policy effect.

3.2. Variable Selections

3.2.1. Corporate Pollution Emission Levels

This paper selects the emission volume of sulfur dioxide, transformed by taking the logarithm, as the measurement index of corporate pollution emission levels, mainly based on three considerations: first, China’s energy consumption structure, which is primarily coal-based, makes atmospheric pollution, especially sulfur dioxide, one of the main pollution indicators; second, industrial sulfur dioxide emissions have become the most significant source of sulfur dioxide emissions in China; third, as sulfur dioxide is the most direct air pollutant perceptible to the public, multiple emission reduction target policies have identified it as the primary pollutant. Additionally, in the pilot of emissions trading, sulfur dioxide is explicitly stipulated as the main pollutant for total quantity control.

3.2.2. Level of Service Integration in Manufacturing Enterprises

Based on the relationship with the manufacturing value chain and drawing on the concept of service relevance, which refers to the degree of consistency or closeness between the services provided by manufacturing enterprises and their main products [36], this paper refines all the service businesses involved by manufacturing enterprises into 17 categories. Among them, the first eight types of services are service businesses based on the extension of the manufacturing value chain, including product installation and after-sales maintenance, remote monitoring and testing, software and information technology solutions, etc., which are defined as embedded services; the last nine types of services are unrelated to the value chain, including trade distribution, logistics, finance, real estate, hotel and catering, tourism services, etc., and are called mixed services. This paper uses the proportion of service business income to operating income to measure the level of service integration of manufacturing enterprises.

3.2.3. Level of Green Innovation in Manufacturing Enterprises

This paper uses the logarithm of the number of green patent applications as the core metric for green innovation activities in the manufacturing industry and uses the logarithm of the number of green invention patent applications to represent high-quality green innovation of enterprises [59,60]. Based on the green patent classification list released by the World Intellectual Property Organization (WIPO), seven major categories of green patents involved by enterprises are selected. The reason is that, compared with the high failure rate and strong uncertainty of R&D investment, innovation output can be regarded as the ultimate embodiment of the input of innovative resources and the efficiency of innovation implementation, which can more intuitively reflect the level of enterprise innovation. At the same time, considering that the process of obtaining patent authorization for enterprises is cumbersome and susceptible to external interference, it is more reasonable to use the number of patent applications, which has more stability and timeliness, to represent the level of green innovation output. In addition, compared with the date of authorization, patent applications can be less affected by factors of corruption and uncertainty in the testing process and can better reflect the actual level of enterprise innovation [61]. Therefore, this paper regards the year of patent application as the year of green innovation output.

3.2.4. Independent Variable

The independent variable in the difference-in-differences model is the pilot policy of emission trading (Treat × Post). Here, Treat is a group dummy variable, taking the value of 1 when the sample manufacturing enterprise is in the “emission trading” pilot area; otherwise, it is 0. Here, Treat is a dummy variable for the group, taking the value of 1 when the sample manufacturing enterprise is in the “pollution rights trading” pilot area; otherwise, it is 0. Specifically, in the baseline regression, if the listed company is located in one of the eleven provinces or cities that were piloting sulfur dioxide pollution rights trading in 2007, its group variable is set to 1; otherwise, it is 0. Post represents the time dummy variable indicating the policy shock, which is 0 before the “pollution rights trading” pilot and 1 afterward. Treat × Post represents the interaction term between the group and time dummy variables, and its coefficient is the policy effect of “pollution rights trading” estimated by the difference-in-differences method. The descriptive statistical results for specific variables are presented in

Table 1. Descriptive statistics.

Variable	Definition	N	Mean	SD	Min	Max
Sulfur Dioxide	Natural logarithm of sulfur dioxide	2438	8.520	5.479	0	17.70
(SO2)	emissions plus one
Independent	Treat × Post	18,870	0.296	0.457	0	1
Variable
Embedded	Embedded service revenue/operating	18,870	0.0257	0.0933	0	1
Services	revenue
Mixed Services	Mixed service revenue/operating	18,870	0.0498	0.108	0	1
	revenue
Green	Natural logarithm of green patent	18,768	0.242	0.668	0	6.441
Innovation	applications plus one
Green	Natural logarithm of green invention	18,768	0.162	0.536	0	6.112
Invention	patent applications plus one
Firm Size	Natural logarithm of total assets	16,527	21.59	1.355	16.70	27.39
Debt-to-Asset
	Total liabilities/total assets	16,516	48.86	121.8	0.752	9696
Ratio
Capital Ratio	Total assets/number of employees	15,641	1,854,895	1,132,962	0	361,275,840
Equity	Major shareholder ownership percentage	14,584	36.41	15.55	3	98.86
Concentration Investment-to-Asset Ratio	Cash paid for the acquisition of fixed assets, intangible assets, and other long- term assets/total assets	16,438	0.0641	0.0592	0	0.642
Profitability	Operating revenue growth rate	15,987	23.37	261.1	−100	25,180

Source: Authors’ calculation based on China Industrial Enterprise Pollution Database, CSMAR and WIND Database.

.

3.3. Data Source

This empirical study primarily involves four databases: the China Industrial Enterprise Pollution Database, the Listed Company Patent Database, the Wind (WIND) Information Financial Terminal, and the Guotai’an (CSMAR) Database. Among them, the China Industrial Enterprise Pollution Database, as a unique dataset, contains the most detailed environmental statistics in China, recording in detail data related to the emission of pollutants by industrial enterprises and their pollution control actions, which has not yet been widely used in academic research. The patent data of listed companies come from the State Intellectual Property Office of the People’s Republic of China (SIPO); The Wind (WIND) Information Financial Terminal and the Guotai’an (CSMAR) Database provide more comprehensive financial data of listed companies as well as various business operation data. This study uses listed manufacturing companies on China’s Shanghai and Shenzhen A-shares for the period 2002–2018 as the initial sample for four interrelated reasons. First, the Shanghai and Shenzhen stock exchanges together represent the full mainland A-share market, covering firms across a broad manufacturing base and spanning both state-owned and private enterprises; as documented, China’s A-shares on these exchanges provide the most comprehensive and representative sample of Chinese manufacturing firms.

Second, the year 2002 marks the earliest point at which consistent firm-level pollution data (from the China Industrial Enterprise Pollution Database) and comprehensive listed-company financial/service–business data (from Wind/CSMAR) are reliably available—thereby ensuring data completeness and allowing for a robust pre-policy period for testing the parallel-trend assumption. Third, the national SO₂ emissions-trading pilot was formally launched in 2007 across 11 provinces/cities (many of which host firms listed on SSE or SZSE), so by selecting 2002 as the start year, we provide at least a five-year pre-treatment window and thus enhance the credibility of our difference-in-differences design. Fourth, the end year 2018 is chosen because by that time the pilot period had matured but preceded major structural reforms and the rollout of a nationwide carbon market—thus capturing medium-term policy effects while avoiding contamination from overlapping reforms. The overall 17-year window (2002–2018) is therefore long enough to observe firm responses in manufacturing, yet constrained enough to maintain identification validity.

Sample inclusion and exclusion criteria further strengthen internal validity. Firms are required to be continuously listed on either SSE or SZSE during 2002–2018, to be classified as manufacturing (per two-digit industry codes), to have no “ST” (Special Treatment) status in any year (to exclude financially distressed or data-irregular firms), and to have no missing core variables (pollution emissions, fixed-asset investment, service-business metrics), thereby forming a balanced panel. Firms are excluded if they experience delisting, major restructuring/merger, or missing key observations for more than three years. This rigorous sample construction helps control for firm fixed effects and selection bias and enhances the credibility of our DID identification strategy. In the empirical testing process, it is necessary to merge the relevant data from the China Industrial Enterprise Pollution Database, the listed company patent data, the Wind (WIND) Information Financial Terminal, and the Guotai’an (CSMAR) Database. Following the matching algorithm described in Brandt et al. (2012), the enterprise pollution database is processed to form a panel data set of industrial enterprise environmental pollution [62]. A similar method is used to construct panel data for the sample of listed manufacturing companies. Subsequently, the enterprise pollution data is merged with the economic data based on the organization code in the pollution emission database and the security code in the listed company data. Our sample inclusion criteria require firms that are continuously listed during 2002–2018, classified as manufacturing (per two-digit industry codes), not labeled ST (Special Treatment) at any point in the sample period (to eliminate firms with significant distress or data irregularities), and with non-missing core variables (pollution emissions, fixed-asset investment, service-business ratio) in all years, thus forming a balanced panel framework. Firms are excluded if they: (i) undergo delisting, major restructuring, or merger during 2002–2018; (ii) have missing or irregular data for key variables for more than 3 years; or (iii) are classified as ST/ST* at any time during the sample. This careful sample construction enhances internal validity, controls for firm fixed-effects and selection bias, and strengthens the credibility of our difference-in-differences identification strategy.

4. Empirical Results

This paper first examines whether the implementation of the pilot emission trading policy has reduced the pollution emissions of enterprises in the pilot areas, and on this basis, further investigates the potential role of the pilot emission trading policy in the selection of transformation and upgrading paths in the manufacturing industry.

4.1. The Effectiveness Test of Environmental Governance Through Emission Trading

4.1.1. Results of the Difference-in-Differences Regression

Using enterprise-level pollution data, we examine the causal impact of China’s SO₂ emissions-trading pilot.

Table 2. The impact of emission trading policies on enterprises’ SO₂ emissions.

VARIABLES	(1) SO2	(2) SO2	(3) Treat × Post	(4) SO2
Treat × Post	−1.808 *** (0.293)	−1.887 *** (0.292)	−0.013 *** (0.001)	−5.043 *** (0.820)
Gf × Post
Iv × Post
Controls	No	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Observations	2438	2436	12,558	2119
R-squared	0.063	0.079	0.388	0.005
F-teat			319.82

Note: The standard error appears in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01. Source: Authors’ calculation based on China Industrial Enterprise Pollution Database, CSMAR, and WIND Database.

reports our difference-in-differences (DID) estimation results. Columns (1) and (2) present two-way fixed-effects regressions, both before and after firm-level matching. In both cases, the coefficient on the interaction variable Treat × Post is significantly negative, indicating that firms in pilot regions experienced a statistically meaningful reduction in SO₂ emissions compared to control firms following policy introduction. We define Treat as manufacturing firms located in the designated pilot provinces/cities, and Post as the period after policy implementation. To bolster identification validity, we conduct a parallel-trend test and placebo regressions, supporting the assumption that pre-policy emission trends did not differ systematically between treatment and control groups. Standard errors are clustered at the province-year level to account for regional policy implementation timing and serial correlation.

To ensure the consistency of the difference-in-differences estimation results, the treatment group and the control group must meet the parallel trends assumption, meaning that the outcome variable maintains a consistent development trend in both groups before policy intervention.

Figure 1 shows the results of the parallel trends and dynamic effects test for the difference-in-differences estimation, indicating that there is no significant difference between the treatment and control groups before the implementation of the pilot emission trading policy, which satisfies the parallel trends assumption.

4.1.2. Instrumental Variable Regression Results

This paper attempts to use reasonable instrumental variables to address the endogeneity problem in the selection of the experimental group. Following the approach of Hering and Poncet (2014), the paper uses the air ventilation coefficient as an instrumental variable for whether a city is included in the sulfur dioxide emission trading pilot program [63]. Specifically, the diffusion of pollution depends on two meteorological factors: the 10 m wind speed (horizontal dispersion) and the boundary layer height (vertical dispersion). Based on the ERA dataset provided by the European Centre for Medium-Range Weather Forecasts and the latitude and longitude data of Chinese prefecture-level cities, the paper calculates the air ventilation coefficient based on the product of unit wind speed and boundary layer height.

Theoretically, using the air ventilation coefficient as an instrumental variable for whether a city is included in the sulfur dioxide emission trading pilot program satisfies the two conditions for an effective instrumental variable: (1) China’s “emission trading” system is a market-based environmental regulation measure based on air pollution control, and its main objective is to control the pollution emissions of enterprises. The smaller the air ventilation coefficient of a region, the higher the monitored concentration of pollutants, and the more likely it is to become a target area for the emission trading pilot program, thus satisfying the relevance assumption of the instrumental variable. (2) The air ventilation coefficient is determined by geographical and meteorological conditions, and it does not directly affect the degree of enterprises’ response to the emission trading policy, thus satisfying the exogeneity assumption of the instrumental variable.

Table 2 columns (3) and (4) report the regression results using the instrumental variable method. The first-stage estimation results show that the coefficient of the interaction term Gf × Post is significantly negative at the 1% level, and the F-statistic is greater than the critical value of 10, indicating that the instrumental variable meets the relevance condition. The second-stage estimation results indicate that the coefficient of the interaction term Iv × Post, which is the focus of attention, remains significantly negative at the 1% level, consistent with the baseline regression. This suggests that after addressing the endogeneity issues related to the selection of the experimental group and omitted variable bias, the emission trading policy can still significantly reduce the SO₂ emission levels of enterprises, and the research conclusion remains robust.

4.2. Examination of the Upgrading and Transformation Effects of Emission Trading

Based on theoretical mechanism analysis, the pilot policy of emission trading, as an important means of market-oriented environmental regulation, significantly influences the transformation of corporate business structure. This paper considers the transformation of the manufacturing industry from two aspects: service integration and green innovation.

Table 3. The impact of emission trading on the service transformation of manufacturing enterprises.

VARIABLES	(1) emb	(2) hyb	(3) lnapply	(4) lnfm
Treat × Post	0.022 *** (0.003)	−0.017 *** (0.004)	−0.010 (0.022)	−0.015 (0.019)
Controls	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Observations	14,525	14,525	14,516	14,516
R-squared	0.033	0.066	0.101	0.090

Note: the standard error appears in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01. Source: Authors’ calculation based on China Industrial Enterprise Pollution Database, CSMAR, and WIND Database.

columns (1) and (2) report the DID estimation results of the impact of the emission trading policy on the service integration of the manufacturing industry, with column (1) showing the regression results of embedded service integration and column (2) showing the regression results of mixed service integration. The results show that the pilot policy of emission trading significantly enhanced the level of embedded service integration in manufacturing enterprises, while significantly reducing the level of mixed service integration. This indicates that the implementation of the emission trading system has led to an optimization effect in services, promoting manufacturing enterprises to engage in embedded services closely related to their main business and to divest mixed services that are not closely related to their main business, thus facilitating the extension of manufacturing enterprises towards the downstream of the value chain.

Table 3 columns (3) and (4) report the DID estimation results of the impact of the emission trading policy on the level of green innovation in the manufacturing industry. Column (3) presents the regression results for green patent applications, while column (4) presents the regression results for green invention patents. The results show that although the coefficients of the key variable Treat × Post are all negative, they are not significant, indicating that the pilot policy of emission trading has no significant impact on either the green innovation activities of enterprises or the high-quality green invention innovation activities. The possible reason is that the increase in pollution control costs has squeezed out non-substantive green innovations that do not fundamentally contribute to enterprise emission reduction and competitiveness enhancement, which has largely alleviated the issue of innovation bubbles. At the same time, green innovation is characterized by high investment and high risk, and the effects of innovation activities are manifested with a certain time lag. The incentive effect of emission trading on high-quality green innovation has not been evident in the short term, showing a non-significant compensatory effect of emission trading on innovation. Therefore, facing the emission trading system, manufacturing enterprises have not “advanced in the face of difficulties” by reducing pollution emissions through green innovation.

4.3. Robustness Test

To ensure the reliability of the empirical results, robustness tests are conducted below by replacing the dependent variable, eliminating interference from other policies, and conducting placebo tests.

4.3.1. Replace the Dependent Variable

This study attempts to re-estimate using alternative measures of service integration and green innovation. Columns (1) and (2) of

Table 4. Robustness test 1.

VARIABLES	(1) embr	(2) hybr	(3) f.fmsy	(4) f.fmshouquan
Treat × Post	0.014 *** (0.003)	−0.015 *** (0.003)	−0.206 (0.215)	−0.013 (0.104)
Controls	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Observations	14,525	14,525	13,430	13,430
R-squared	0.013	0.047	0.031	0.023

Note: the standard error appears in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01. Source: Authors’ calculation based on China Industrial Enterprise Pollution Database, CSMAR, and WIND Database.

re-examine the level of service integration using the ratio of enterprise service revenue to total assets, and find that the emission trading program significantly increased the level of embedded services in manufacturing enterprises, while significantly decreasing the level of mixed services. For green innovation, as it often takes 1–2 years for a patent to be applied for and granted, this study uses the number of green patents and green invention patents granted with a one-year lag as alternative variables for green innovation output and high-quality green innovation output to conduct a robustness check. As shown in columns (3) and (4) of Table 4, while the emission trading program reduced the level of green patents and green invention patents held by enterprises, the regression results are not statistically significant.

4.3.2. Change the Estimation Model

This study further employs the propensity score matching (PSM) method to address the endogeneity problem caused by potential sample selection bias at the enterprise level. The basic logic of this method is to synthesize various characteristic variables of the enterprises into a propensity score and then match the treatment group and control group samples based on the proximity of the propensity scores. This ensures that the empirical subjects better meet the parallel trend assumption required for the difference-in-differences (DID) estimation.

Subsequently, the panel DID estimation is conducted using the successfully matched samples, which can attenuate the systematic differences between enterprises and reduce the estimation bias of using the DID method alone.

Table 5. Robustness test 2.

VARIABLES	(1) emb	(2) hyb	(3) lnapply	(4) lnfm
Treat × Post	0.021 *** (0.003)	−0.016 *** (0.004)	−0.007 (0.022)	−0.012 (0.019)
Controls	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Observations	14,525	14,525	13,430	13,430
R-squared	0.013	0.047	0.031	0.023

Note: the standard error appears in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01. Source: Authors’ calculation based on China Industrial Enterprise Pollution Database, CSMAR, and WIND Database.

reports the PSM-DID estimation results on the impact of the emission trading policy on the transformation and upgrading of the manufacturing industry, and it is found that the estimation results do not change substantially, whether for service integration or green innovation.

4.3.3. Exclude Interference from Other Policies

Facing the increasingly severe international climate and environmental issues, China has implemented various pollution control measures, such as the “Two Control Zones” policy (acid rain control zone or SO₂ control zone) launched in 2000, the “Regulation on the Collection and Use of Pollutant Discharge Fees” proposed in 2002, and the specific emission reduction requirements of 10% and 8% reduction in total SO₂ emissions proposed in the “11th Five-Year Plan” and “12th Five-Year Plan”, which may all affect the difference-in-differences (DID) estimation results. Based on this, this study excludes the impact of the pollutant discharge fee collection policy, the emission reduction requirements of the “11th Five-Year Plan” and “12th Five-Year Plan”, as well as the “Two Control Zones” policy. First, the regression model controls for the impact of the amount of pollutant discharge fees in each province by adding pollutant discharge fee collection as a control variable. Second, time dummy variables are added for the years 2006 and after, and the data sample for 2011 and after is deleted in order to control for the policy impact of the “11th Five-Year Plan” and “12th Five-Year Plan”. Third, the treatment group and control group used for the DID estimation are reconstructed to further exclude the impact of the “Two Control Zones” policy, i.e., the regions that implemented both the emission trading pilot and the “Two Control Zones” policy are treated as the treatment group, while the regions that only implemented the “Two Control Zones” policy are treated as the control group. By comparing the changes in SO₂ emissions, service integration, and green innovation before and after the policy, the impact of the “Two Control Zones” policy can be well excluded.

Table 6. Robustness test 3.

VARIABLES	(1) emb	(2) hyb	(3) lnapply	(4) lnfm
Treat × Post	0.008 ** (0.003)	−0.010 ** (0.005)	0.020 (0.019)	0.006 (0.015)
Lnfee	−0.003 (0.003)	−0.004 (0.004)	−0.042 ** (0.017)	−0.055 *** (0.012)
post2006	−0.003 (0.003)	−0.004 (0.004)	−0.042 ** (0.017)	−0.055 *** (0.012)
Controls	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Observations	14,525	14,525	13,430	13,430
R-squared	0.013	0.047	0.031	0.023

Note: the standard error appears in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01. Source: authors’ calculation based on China Industrial Enterprise Pollution Database, CSMAR, and WIND Database.

reports the net effect of the SO₂ emission trading policy after excluding the interference of other policies, and it is found that the estimation results are consistent with the previous regression results, indicating that the empirical conclusions remain robust after excluding policy interference.

4.3.4. Placebo Test

To further ensure the robustness of the research conclusions, this study adopts the “counterfactual” method to verify whether the treatment group and the control group have a common trend. The basic principle and process are: assuming that the “emission trading” pilot is advanced to 2006, the corresponding dummy variables are constructed to perform the difference-in-differences (DID) regression. If the coefficients of the policy and time interaction terms are not significant in the false policy setting, it indicates that the treatment group and the control group have a common trend, confirming that the changes in emission reduction and transformation effects are caused by the emission trading system rather than other factors. Otherwise, it suggests that the conclusion is not robust. Furthermore, to eliminate the interference of the emission trading pilot, the sample period is controlled between 2002 and 2006, and the regression results are shown in

Table 7. Robustness test 4.

VARIABLES	(1) emb	(2) hyb	(3) lnapply	(4) lnfm
Treat × Post	0.016 (0.010)	−0.004 (0.008)	0.018 (0.021)	0.002 (0.015)
Controls	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Observations	14,525	14,525	13,430	13,430
R-squared	0.013	0.047	0.031	0.023

Note: The standard error appears in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01. Source: Authors’ calculation based on China Industrial Enterprise Pollution Database, CSMAR, and WIND Database.

. It is easy to find that under the assumption of virtual policies at different time points, the coefficients of the interaction terms are not all significant, indicating that the treatment group and the control group obtained after PSM meet the conditions for using the difference-in-differences method and have common trend characteristics. This also reflects that the changes in the level of emission reduction and transformation effects in the manufacturing industry are indeed caused by the emission trading policy.

5. Further Research

The emission trading system has reduced the sulfur dioxide emissions of enterprises in the pilot regions while generating a service optimization effect, but the innovation incentive effect has not been realized.

5.1. Analysis of Upgrading and Transformation Mechanism

5.1.1. Service Optimization Mechanism

To examine the intermediary mechanism of the service optimization effect, first, the proportion of sales to the top five customers is used to measure the customer concentration of manufacturing enterprises. The higher the customer concentration, the easier it is for the enterprise and its customers to form a stable and reliable customer relationship, which helps to enhance the willingness, frequency, and depth of sharing customer demand information between the two parties, thereby enhancing the enterprise’s willingness and ability to undergo service transformation [64]. Second, the proportion of fixed asset investment is used to measure the capital allocation structure of the enterprise, and a higher level of fixed asset investment can provide the corresponding material basis for the transformation and upgrading of product-related services of manufacturing enterprises. The results of the mechanism test (see

Table 8. Mechanism examination of service optimization effects.

VARIABLES	(1) Customer Concentration	(2) Fixed Assets	(3) Current Assets	(4) Marketing Intensity
Treat × Post	2.109 *** (0.531)	0.014 *** (0.004)	−0.003 * (0.001)	−0.005 *** (0.002)
Controls	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Observations	13,345	14,519	14,498	14,525
R-squared	0.029	0.072	0.039	0.316

Note: The standard error appears in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01. Source: Authors’ calculation based on China Industrial Enterprise Pollution Database, CSMAR, and WIND Database.

) show that the coefficients of customer concentration and fixed assets are both significantly positive at the 1% level. On the one hand, this indicates that the implementation of the emission trading system has forced enterprises to pay more attention to the downstream customer demand, resulting in a significant increase in customer concentration, which in turn promotes the increase in service business. On the other hand, the emission trading system can increase the fixed asset investment of enterprises in the pilot regions, increase the production and operation capabilities of enterprises, and provide the corresponding material basis for the transformation and upgrading of product-based services, thereby promoting the increase in embedded services.

Unlike the embedded services oriented towards value chain extension, mixed-in services have a relatively small connection with the core manufacturing business. The development of such businesses requires the consumption of more working capital and sales expenses in order to explore unfamiliar business areas. This paper uses the proportion of current assets and marketing intensity to reflect the tendency of manufacturing enterprises to invest in mixed-in services and examines the intermediary mechanism of the impact of emission trading on mixed-in services. The results show that both the proportion of current assets and marketing intensity are significantly negative, indicating that the emission trading system has strengthened the motivation of enterprises to control pollution, reduced the investment in liquid assets and marketing expenses, and thus prompted manufacturing enterprises to shed some mixed-in services without resource advantages.

Environmental regulation tools, the emission trading system, based on clear market price signals, fully ensure the flexibility of enterprises in the emission reduction process. After the implementation of the emission trading system, market-oriented environmental regulation allows enterprises to fulfill their commitments while minimizing the cost of emission reduction. Considering the protection of long-term economic interests, manufacturing enterprises will reallocate production factors and resources, reducing investment in inefficient and highly polluting production departments while increasing resources in high-efficiency and clean production and operation departments. In this way, the increase in customer concentration provides greater motivation and financial support for the green service transformation, prompting enterprises to increase fixed asset investment and, through the optimization of service integration, extend their business activities downstream in the value chain, achieving a “splendid” transformation of manufacturing enterprises.

5.1.2. Innovation Incentive Mechanism

Facing the emission reduction pressure under the emission trading system, R&D investment and end-of-pipe treatment are two feasible choices, and there is a certain substitution relationship between them. Within the scope of moderate emission reduction, end-of-pipe treatment can achieve the purpose of reducing pollution emissions by relying on R&D investment to obtain green patents. This paper chooses the number of desulfurization facilities of enterprises to represent the level of pollution control investment [65] and the desulfurization capacity of desulfurization facilities to measure the pollution control level of enterprises [66]. The R&D investment intensity of enterprises is used as an indicator of green innovation investment, and the regression results are in

Table 9. Examination of the mechanism behind the innovation incentive effect.

VARIABLES	(1) Desulfurization Facilities	(2) Desulfurization Capability	(3) R&D Investment
Treat × Post	0.090 ** (0.039)	0.337 ** (0.146)	0.002 (0.053)
Controls	Yes	Yes	Yes
Year	Yes	Yes	Yes
Observations	2526	2526	9872
R-squared	0.169	0.136	0.010

Note: the standard error appears in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01. Source: authors’ calculation based on China Industrial Enterprise Pollution Database, CSMAR, and WIND Database.

are obtained using the difference-in-difference method. The results of the mechanism test show that the coefficients of desulfurization facilities and desulfurization capacity are both significantly positive, while the coefficient of R&D investment is positive but not significant, indicating that the emission trading system has driven enterprises to reallocate production factors, increased the level of end-of-pipe pollution control investment, and focused on enhancing pollution control capabilities [16]. However, under the constraint of limited resources, it has not been able to significantly increase R&D investment, which is detrimental to the output of green patent achievements.

Why do manufacturing enterprises not “forge ahead in the face of difficulty” and rely on green innovation to achieve transformation and upgrading? The reason is that investments in end-of-pipe pollution control have clear effects, and the installation of desulfurization facilities and the increase in desulfurization capacity can reduce sulfur dioxide emissions with certainty. In contrast, although R&D investments can fundamentally solve pollution emission problems, there is a high degree of uncertainty, and the effects cannot be immediately apparent. After the implementation of the emissions trading system, rational manufacturing enterprises install more end-of-pipe emission reduction equipment to achieve the goal of reducing sulfur dioxide emissions. However, the increase in pollution control investments crowds out investments in innovation factors, which suppresses innovation activities in the upstream of the manufacturing industry value chain.

5.2. Heterogeneity Enterprise Perspective on the Differences in Effects

The previous discussion explored the impact of emissions trading on the transformation and upgrading of enterprises. Then, what factors ultimately influence an enterprise’s choice of service optimization and green innovation? After the implementation of the emissions trading system, what types of enterprises can achieve service optimization and green innovation upgrades? Next, this paper examines the moderating effect of manufacturing enterprises’ service optimization from the aspects of production chain position and digitalization level and tests the conditions for the occurrence of green innovation incentives in manufacturing enterprises from the aspects of market power and technological foundation.

5.2.1. Moderating Effects of Regional and Industry Characteristics on Stakeholder Responses

The different positions of enterprises in the production chain lead to differences in the product division of labor, thereby affecting the service optimization and upgrade of manufacturing enterprises. This paper uses the industry production chain position adjusted based on the business scope of listed companies to measure the position of manufacturing enterprises in the production chain. The larger the index, the closer the enterprise’s manufacturing products are to the production starting point based on raw materials; conversely, the closer to the production and manufacturing of the final product.

Table 10. Examination of the mechanism behind the innovation-incentive effect: Production chain position and digitalization embedding.

VARIABLES	(1) Position in the Production Chain emb	(2) hyb	(3) Level of Digitalization emb	(4) hyb
Treat × Post × Var	0.013 *** (0.003)	−0.001 (0.002)	0.003 * (0.002)	−0.003 (0.002)
Treat × Post	−0.016 (0.010)	−0.018 * (0.010)	0.020 *** (0.004)	−0.015 *** (0.004)
Var	-	-	0.005 *** (0.001)	0.003 ** (0.001)
Controls	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Observations	8213	8213	14,384	14,384
R-squared	0.048	0.057	0.034	0.066

Note: the standard error appears in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01. Source: authors’ calculation based on China Industrial Enterprise Pollution Database, CSMAR, and WIND Database.

, columns (1) and (2), presents the test results for the moderating effect of the production chain position. It can be seen that for embedded services, the estimated coefficient of the DID interaction term with the production chain position is significantly positive, indicating that for upstream enterprises in the production chain, emissions trading significantly enhances the service optimization upgrade in manufacturing. The reason is that after the implementation of emissions trading, upstream enterprises in the production chain can better utilize market-oriented environmental regulations to improve the single production mode of enterprise products, fully leverage the competitive advantage of service integration, and ultimately achieve service optimization upgrade.

Considering the rapid development of the digital economy has provided a favorable opportunity for the service optimization upgrade and value chain climbing of the manufacturing industry [67]. This paper uses text mining techniques to construct a digitalization index at the enterprise level, using the logarithm of the frequency of digitalization-related keywords in the annual reports of listed companies, and further examines the differences in the transformation and upgrading effects of emissions trading on enterprises with different levels of digitalization. The results of the examination of the moderating effect of digitalization level in columns (3) and (4) of Table 10 show that the estimated coefficient of the interaction between DID and market power is significantly positive for embedded services, while the interaction coefficient is negative for hybrid services. This indicates that the implementation of the emissions trading policy is beneficial for the service optimization and upgrading of enterprises with high digitalization levels, further promoting manufacturing enterprises to engage in embedded services closely related to their core business, and helping manufacturing enterprises maintain their competitive advantage through service optimization, achieving the upgrading effect of extending downstream in the value chain.

5.2.2. Interaction Mechanisms Linking Carbon Disclosure and Stakeholder Behavior

Given the differences in the transformation decisions of enterprises with varying pricing power, this study uses the Lerner index as an indicator to measure the market power of manufacturing enterprises and then constructs an interaction term with policy effects to reflect the moderating effect of market power on emission trading and green innovation. The results of the test on the moderating effect of market power are shown in columns (1) and (2) of

Table 11. Examination of the conditions for the occurrence of innovation incentive effects.

VARIABLES	(1) Market Power lnapply	(2) lnfm	(3) Technological Foundation lnapply	(4) lnfm
Treat × Post × Var	−0.398 *** (0.088)	−0.271 *** (0.074)	0.021 *** (0.007)	0.015 ** (0.006)
Treat × Post	0.084 *** (0.030)	0.049 ** (0.025)	−0.075 ** (0.032)	−0.061 ** (0.027)
Var	0.042 (0.062)	0.015 (0.052)	0.169 *** (0.006)	0.116 *** (0.005)
Controls	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Observations	14,511	14,511	13,430	13,430
R-squared	0.102	0.091	0.178	0.143

Note: the standard error appears in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01. Source: authors’ calculation based on China Industrial Enterprise Pollution Database, CSMAR, and WIND Database.

. It can be seen that in the case of green innovation, the estimated coefficient of the interaction term between DID and market power is significantly negative, indicating that for enterprises with high market power, emission trading significantly suppresses the green innovation upgrade of the manufacturing industry, and the enterprises’ motivation for upstream green innovation upgrade is insufficient. Additionally, in the case of green innovation, the estimated coefficient of the interaction term between DID and market power is negative but not significant, indicating that market power cannot significantly affect the high-quality green innovation level of enterprises. The reason for this is that for enterprises with a certain market power, faced with the compliance costs brought about by the implementation of the emission trading system, enterprises are more inclined to choose a cost-shifting strategy rather than a green innovation upgrade strategy. In contrast, for enterprises with low market power in a competitive environment, they have higher production and operation efficiency, and the implementation of market-based environmental regulations may be more conducive to stimulating their motivation for green innovation. Therefore, the difference in market power is an important reason why the green innovation incentive effect of emission trading is difficult to achieve.

Considering the differences in the existing technological foundations of manufacturing enterprises, the choices in responding to environmental regulations vary.

This study uses the stock of patents of manufacturing enterprises to characterize the technological foundation conditions of enterprises. The results of the examination of the moderating effect of technological foundation are shown in columns (3) and (4) of Table 11. It can be found that the estimated coefficients of the interaction between DID and technological foundation are significantly positive, regardless of green innovation or green invention innovation. This means that compared to enterprises with weak technological foundations, manufacturing enterprises with better technological foundations have advantages in transforming and upgrading towards green innovation. After the implementation of emission trading, enterprises with high technological foundations, relying on their strong technological capabilities and previous accumulation, have significantly increased their green innovation output. The possible reason is that enterprises with high technological foundations are able to cope with the large upfront investments required for innovation activities, have higher operational efficiency, and have comparative advantages in overcoming technological thresholds. Therefore, as the technological foundation of enterprises continues to improve, the implementation of market-based environmental regulations is more conducive to stimulating the motivation of enterprises to engage in green innovation.

6. Discussion

This study contributes to the literature on market-based environmental regulation and firm transformation by examining how China’s SO₂ emission trading system influences manufacturing firms’ upgrading behavior through a value-chain perspective. Our findings reveal three distinctive patterns.

First, consistent with policy expectations of emission trading frameworks, the system yields a significant reduction in sulfur dioxide emissions and a synergistic effect on other pollutants, indicating that market mechanisms can complement traditional command-and-control regulation to achieve environmental gains. However, the transformation path of manufacturing firms was not uniformly green innovation-oriented. Instead, firms predominantly pursued a service optimization pathway, characterized by increased embedded services downstream and a reduction in mixed/non-core services. This suggests that firms facing regulatory pressure preferred value-chain extension into services rather than upstream technological innovation.

Second, our empirical evidence shows that the emission trading system did not significantly increase R&D investments or breakthrough green innovation among the sampled firms. Mechanism tests indicate that firms relied on end-of-pipe pollution control, increased customer concentration, and fixed asset investments as the conduit for service optimization, rather than enhancing stream investments in R&D. This finding implies that while the system induces environmental compliance, it does not automatically trigger the ‘Porter effect’ of innovation unless complementary institutional conditions are met—such as absorptive capacity or subsidies.

Third, heterogeneity analysis reveals that embedded service upgrading among upstream firms and higher digitalization levels strengthen the service-oriented upgrading effect, whereas large “high-power” enterprises with cost-shifting advantages face stronger hurdles for green innovation upgrading upstream in the value chain. This underscores that structural position, market power, and internal capabilities shape firm responses to environmental regulation.

Collectively, these results underscore that market-based environmental regulation can be effective in driving pollution reduction and service-oriented upgrading but may fall short of prompting structural technological innovation unless supported by targeted governance measures. For policymakers, this implies that emission trading systems need to be complemented by mechanisms that stimulate R&D, strengthen firms’ innovation capabilities, and mitigate the inertia of cost-driven compliance behaviors. Future research should explore the long-run trajectories of service upgrade versus innovation uptake and investigate how firm-level heterogeneity and institutional contexts moderate these pathways.

7. Conclusions and Policy Recommendations

In order to fully implement the strategy of sustainable development and accelerate the promotion of green and low-carbon growth, it is imperative to strengthen the coordinated mechanism of ecological civilization, fully leverage the decisive role of market mechanisms, and improve the ecological and environmental governance system. This study takes the SO₂ emissions trading system (ETS) as the empirical focus and examines how such market-based environmental regulation drives the transformation and upgrading of manufacturing firms through a value-chain perspective—specifically by identifying a downstream service-extension (service-optimization) pathway and an upstream innovation-incentive pathway. Utilizing the 2007 pilot SO₂ ETS in 11 Chinese provinces and listed manufacturing firms in China from 2002 to 2018 via a difference-in-differences (DID) method, our main findings are as follows: firstly, the ETS significantly reduced manufacturing firms’ sulfur dioxide emissions and generated a synergistic emission-reduction effect; secondly, the ETS induced service-oriented upgrading in firms, manifested by a significant increase in embedded services aligned with value-chain extension and a significant decrease in mixed, non-core services; thirdly, the ETS did not significantly enhance firms’ green innovation levels—firms mainly relied on end-of-pipe pollution-control investment rather than R&D investment; fourthly, moderating analyses show that firms upstream in the production chain and with higher levels of digitalization responded more for service extension, whereas high-power firms with cost-shifting advantages demonstrated weaker green innovation upgrading upstream in the value chain.

Based on these findings, we derive the following three sets of concrete practical recommendations, with specific guidance on modifications to the pilot ETS conditions and prospects for nationwide scale-up:

• Refine pilot emissions-trading conditions and plan national scaling.

Introduce a tiered quota-reduction roadmap in the SO₂ ETS pilot: for example, firms achieving a ≥10% improvement in fixed-asset investment efficiency or embedding ≥20% of service business may qualify for accelerated quota-reduction phases, rewarding service-oriented upgrading rather than basic compliance.

Conduct a nationwide cost–benefit analysis for scale-up: estimate the incremental cost per ton of SO₂ reduced through service upgrading versus technological innovation; based on our micro-evidence, scaling the model from 11 to all 31 provinces could generate additional embedded-service revenues while achieving Y% additional emission reductions. Accordingly, establish transitional subsidies (≈10–15% of compliance cost) to support firms in the early stage of market-driven upgrading.

Enhance governance by establishing a real-time trading dashboard tracking both quota flows and service-conversion metrics (e.g., embedded-service share), and expand cross-regional trading zones gradually—initially an East–Central–West corridor pilot by 2026, with full national integration by 2030.

2.

Promote service-capability optimization as a green manufacturing transformation lever.

Launch a service-transformation incentive fund: provide offset grants covering up to 30% of additional fixed-asset investment or customer-concentration growth aimed at embedded service development. Provide tax credits for manufacturing firms whose revenue from service-enabled value-chain positions exceeds 15% of total revenue.

Set national targets for service extension: for key manufacturing sectors (e.g., equipment, chemicals, textiles), increase the average embedded service ratio by 10% over five years, thereby boosting the service economy, reducing regulatory burdens, and improving firm competitiveness.

Incorporate service business metrics into the ETS compliance framework: from 2027 onward, reporting modules should include the number of joint service-manufacturing projects and the percentage of service revenue, making them part of the compliance assessment.

3.

Activate green-innovation pathways through a dual engine of market and government.

Introduce a green-innovation voucher scheme linked to ETS compliance: firms meeting quota-reduction targets may redeem vouchers (e.g., up to RMB X million) for green R&D or digital-service solutions, aligning compliance incentives with innovation investment.

Establish a national green-innovation fund for manufacturing-ETS participants: prioritize mid-cap firms with intermediate technological capabilities (rather than only high-tech leaders) to avoid innovation elite capture. Allocate an initial fund of RMB Y billion to catalyze broad-based green innovation diffusion.

Implement annual audited reviews of innovation performance: audits should validate firms’ patents and green-product revenues and link these outcomes to ETS benefits, ensuring that subsidy systems complement rather than duplicate market-mechanism incentives.

With the above targeted recommendations, China’s SO₂-emissions-trading pilots can evolve into a scalable, sustainable mechanism that not only reduces pollution but also catalyzes the transformation of manufacturing towards service-rich, technology-enabled, high-value industries, thereby supporting the dual goals of environmental improvement and economic upgrading.

8. Limitations of the Study

While this study advances understanding of how SO₂ emission trading influences manufacturing firms’ transformation and upgrading, several limitations should be acknowledged. First, our sample is limited to listed manufacturing firms in China over the period 2002–2018, which may restrict the generalizability of the findings to non-listed enterprises, other industries, or international contexts. Second, although we find evidence of service-oriented upgrading, our measure of green innovation (e.g., R&D investment) may not fully capture all dimensions of technological upgrading, and the relatively short observation window may underestimate longer-term innovation effects. Third, our categorization of service transformation—into ‘embedded services’ vs. ‘mixed services’—may not comprehensively reflect the full complexity of firms’ service-oriented strategies, and measurement limitations remain. Fourth, although we apply a difference-in-differences design with robustness checks, unobserved heterogeneity in policy implementation across provinces or industries (such as pilot scope, enforcement stringency, or firm responsiveness) may influence the estimated effects.

Future research should consider broader firm samples (including non-listed and small/mid-sized enterprises), longer post-policy observation windows, richer service and innovation metrics, and more granular data on policy enforcement heterogeneity to deepen insight into the mechanisms of market-based environmental regulation.