The Impact of Environmental Subsidies and Enforcement on Green Innovation: Evidence from Heavy-Polluting Enterprises in China

Abstract

This paper investigates the critical role of government policies in promoting green development through enterprise-level green innovation (GI). We specifically examine the impact of environmental subsidies (ES) on heavily polluting enterprises and analyze how government environmental enforcement (EE) moderates this relationship. Furthermore, we explore how these effects vary across different ES characteristics and allocation patterns. Using data from heavy-polluting enterprises listed on China’s A-share market from 2012 to 2021, our analysis reveals several key findings. First, ES has unexpectedly hindered GI in heavily polluting enterprises. This negative effect stems from both ES characteristics and allocation patterns. Specifically, advance subsidies, targeted subsidies, and pollution prevention subsidies prove ineffective. The adverse impact of ES on GI is particularly pronounced in state-owned enterprises, companies lacking environmental protection investments, and firms engaging in earnings management practices. However, our findings indicate that strong environmental enforcement (EE) can effectively mitigate these ES-related issues. Under strict EE, ES demonstrates a positive effect on GI and addresses problems associated with both ES design characteristics and allocation patterns. These results offer valuable insights for government policymakers, particularly in developing countries undergoing green transition, to design and implement more effective environmental policies.

1. Introduction

As China experiences sustained economic growth, the nation faces increasingly severe environmental challenges, particularly in water and air pollution [1,2]. These environmental issues stem primarily from heavy-polluting industries characterized by excessive resource consumption, high emission levels, and substantial environmental degradation [3,4]. Research from the Center for Energy & Environmental Policy Research at the Beijing Institute of Technology reveals that, since China’s reform and opening up, enterprises generating 40% of industrial GDP have consumed approximately 70% of the country’s energy resources. Among 41 national industries, the electricity and thermal production sector, non-metallic mineral industry, and ferrous smelting and calendering industry are the top three contributors to SO₂ and nitrogen oxide emissions. These three industries alone account for more than 60% of SO₂ emissions, 75% of nitrogen oxide emissions, and 70% of smoke and powder emissions from key industrial firms. Therefore, facilitating the green transformation of heavy-polluting industries has become a critical imperative for advancing China’s transition toward a green economy.

For heavy-polluting enterprises in China, green innovation (GI) represents a critical pathway toward sustainable development through the implementation of environmentally friendly practices. GI encompasses scientific and technological innovation activities undertaken by relevant entities to mitigate environmental risks [5,6]. The primary objective of GI is to advance the development of green technologies that enhance energy efficiency, reduce emissions, promote clean production processes, and maximize the utilization of renewable energy sources. These advancements enable firms to enhance their environmental sustainability [6,7,8]. Through GI implementation, heavy-polluting enterprises can achieve multiple benefits, including gaining competitive advantages and long-term sustainable development [9,10].

However, there is currently a lack of GI incentives for heavy-polluting enterprises [11]. GI is marked by long cycles, significant technological uncertainties, and substantial early-stage investments [12]. Furthermore, GI demonstrates dual externalities, including positive environmental impacts and knowledge spillovers comparable to conventional innovations [13,14]. The market mechanism alone proves insufficient to support GI initiatives due to their inherent high risks and substantial capital requirements [15]. This is particularly evident in China, where heavy-polluting enterprises continue to rely predominantly on pollution-intensive technologies during the country’s transition toward a green economy [16]. Statistical evidence underscores this challenge, as more than 65% of heavy-polluting companies in China did not file any green patent applications between 2010 and 2020 [17].

Environmental subsidies (ES) serve as a critical policy instrument for governments to promote GI among heavy-polluting enterprises [18,19,20]. The inherent externalities of GI often create financial barriers for firms, as the private implementation costs typically exceed potential returns [21,22]. ES, provided as unconditional government support, offers enterprises essential external resources to pursue GI initiatives [23,24,25]. While scholarly attention on the relationship between ES and GI has increased, research findings remain inconclusive. Numerous studies demonstrate that ES effectively stimulates GI [19,26,27,28,29], while other research suggests contrary effects [18,30,31].

Similar to ES, environmental enforcement (EE) serves as another crucial policy tool for governments to regulate enterprise environmental protection behavior, and it has been widely implemented globally [32,33]. EE effectively and directly regulates corporate environmental damage behaviors [32]. However, the relationship between EE and corporate GI remains ambiguous. Some studies suggest that, under government EE pressure, enterprises tend to adopt environmentally friendly technologies to meet regulatory requirements [34,35,36]. Conversely, other research indicates that EE’s impact is limited and insufficient to fundamentally transform enterprise production patterns [24,37].

As a developing country undergoing green transformation, China’s ES and EE are experiencing a distinctive evolutionary phase. While the intensity of ES initially increased, it has subsequently shown a consistent declining trend. According to the National Bureau of Statistics of China, fiscal expenditures for environmental protection demonstrated an overall upward trajectory from 2007 to 2023, yet averaged only 2.48% of total fiscal expenditures annually. Furthermore, the China Ecological Environment Statistical Yearbook, published by the Ministry of Ecology and Environment, reveals that, from 2007 to 2022, China’s total social investment in environmental pollution control averaged just 1.25% of GDP, declining to 0.7% in 2022. This stands in stark contrast to Sweden, where the corresponding figure ranged between 2% and 10% from 1998 to 2002. International experience suggests that, when social environmental protection investment reaches 1–1.5% of GDP, it can effectively curb environmental degradation, while a 2–3% investment can actively improve environmental quality. Conversely, China’s EE intensity has been steadily increasing. In the past decade, heavy-polluting enterprises have faced increasingly rigorous environmental regulations, most notably through the implementation of the new environmental protection law in 2015, widely regarded as “the most stringent environmental protection law in history” [38].

In this study, we examine the impact of ES on GI among China’s heavy-polluting listed enterprises from 2012 to 2021, while investigating the moderating role of EE. Our findings reveal that ES negatively affects GI in heavy-polluting enterprises. However, EE serves as a positive moderator, mitigating this negative impact. Through analyzing the timing, form, and content of ES, we discover that the negative effects on GI primarily stem from ES beforehand, targeted ES, and prevention-focused ES initiatives. Furthermore, our research explores how ES impacts GI differently across various enterprise characteristics, including property rights, environmental investment situations, and earnings management levels. The results indicate that ES’s effectiveness is particularly diminished in state-owned enterprises, companies lacking environmental investments, and firms with high earnings management levels. Notably, EE demonstrates its ability to moderate these adverse effects.

Our research significantly expands upon the “narrow” version of the Porter hypothesis, which suggests that specific government policies can stimulate innovation [39,40,41,42]. Firstly, we demonstrate that various design features of ES differently impact GI in heavy-polluting enterprises, providing valuable insights for ES policy design. Secondly, our findings reveal that stringent EE can effectively mitigate the negative effects of ES, highlighting the critical importance of synergy between environmental policies. Lastly, our analysis of different groups based on property rights, environmental investments, and earnings management levels. These findings offer meaningful implications for regional green development during environmental transformation periods.

The following section reviews the existing literature and develops hypotheses. Section 3 describes the sample, data, variables, and research methodology. Section 4 presents and analyzes the research findings. Finally, the concluding section provides policy recommendations to address the identified issues.

2. Literature Review and Hypothesis Development

2.1. The Impact of Environmental Subsidy on Green Innovation

GI refers to scientific and technological innovation activities undertaken by stakeholders to promote environmental protection [5]. For firms, GI exhibits unique dual externalities, environmental positive externality due to its green development objectives, and knowledge spillover effects [13,14,20]. These externalities often make GI activities financially unfavorable for firms, as the private costs typically exceed potential revenues [27]. Consequently, enterprises frequently lack adequate incentives to pursue GI initiatives.

Heavy-polluting enterprises are critical subjects in the GI research literature. These enterprises face significant challenges, including high implementation costs and technical barriers [41]. And, environmental pressures compel heavy-polluting industries to adopt GI practices for sustainable growth [43]. Researchers have extensively investigated various approaches to promote GI within heavy-polluting enterprises [3,15,44]. Moreover, numerous studies have examined the feasibility and effectiveness of policy measures designed to stimulate GI implementation in heavy-polluting enterprises [4,28,41,45].

ES are non-repayable financial support subsidies provided by governments, aimed at promoting environmental protection and governance activities [27,31,46]. These subsidies incentivize enterprises to enhance their products and processing technologies [18,19,29]. The effects of ES remain unclear. Some studies support a clear positive correlation between ES and GI. As supplementary financial resources, ES can help enterprises alleviate operational pressures, enabling them to increase their GI practices [19]. Shao and Chen [27] delved into the impacts of government subsidies, encompassing environmental protection subsidies, R&D subsidies, and talent subsidies, on the green innovation transformation of Chinese enterprises during the period from 2007 to 2019. Their research validated the positive influences of these subsidies. Similarly, Wei et al. [28] utilized the data of 825 A—share listed enterprises from 2009 to 2020 to analyze the correlation between environmental subsidies and GI. Their findings revealed the promotional environmental subsidies on GI.

However, the effectiveness of ES in fostering corporate GI has notable limitations [30]. Some scholars have emphasized that ES is not always an ideal choice for the government, because they only have a positive impact on GI under certain specific circumstances [18,47]. Osorio and Zhang [48] conducted a comparative analysis of product feature subsidies and CO₂ abatement subsidies on corporate environmental investments, revealing that CO₂ abatement subsidies proved beneficial only under conditions of low emission levels. Furthermore, some other scholars believe that ES may hinder corporate GI [30]. Li and Xiao [31] identified that ES can actually impede GI due to companies’ opportunistic behaviors and their tendency to align with government preferences rather than genuine innovation goals.

ES, as a crucial government tool for supporting and guiding enterprises’ green development, theoretically plays a significant role in facilitating the green development of heavily polluting enterprises. However, in China, ES’ effects are limited. First, China’s ES intensity remains inadequate. In 2006, the “Expenditure on Environmental Protection” category-level account was set up in China’s budget system for the first time. According to data from the National Bureau of Statistics, from 2007 to 2023, the proportion of environmental protection expenditure in GDP in China ranged from 0.37% to 0.75%. In developed countries such as the United States, Germany, and Japan, this proportion exceeded 2% as early as the 1970s. Second, ES are broadly targeted at environmental protection activities rather than specifically designed for GI initiatives. Given that GI projects typically involve higher risks, greater costs, and longer payback periods compared to other environmental practices [34]. Finally, the inefficient distribution of ES can lead to distorted allocation of funds both within and between enterprises [31,49,50]. Based on these observations, this article proposes the following hypothesis.

H1. 

ES negatively impacts the GI of heavy-polluting enterprises in China.

2.2. Environmental Subsidy Characteristics and Green Innovation

Various scholars have highlighted the influential effects of environmental policy characteristics on GI [12,51,52]. According to the allocation process and behavioral targets, the design characteristics of ES can be explained from three perspectives [34,53,54]. The first aspect is the subsidy timing [55]. The distinction lies in whether the intended goals are anticipated or achieved, or whether the projects are still ongoing or completed [56]. ES beforehand (ESB) is provided to companies before their project initiation, whereas ES afterwards (ESA) is provided thereafter [47,51,57,58]. The second category is the subsidy form [21,59]. ES can be divided into two categories: general ES (GES) and targeted ES (TES). TES tends to have more explicit guidelines for application, allocation, and investment decisions. GES is ES with no specific goals [21]. The last type of subsidy content. Enterprises have two primary methods of reducing emissions: pollution prevention and end-of-pipe control [60,61]. Correspondingly, ES can be divided into pollution prevention subsidy (PES) and end-of-pipe control subsidy (EES).

ES with different characteristics influences GI differently among China’s heavy-polluting enterprises. Firstly, compared to ESA, ESB provides clearer objectives and more substantial initial funding to guide firms’ GI. However, ESB may introduce challenges absent in ESA, including inequitable distribution, unclear implementation guidelines, and potential subsidy fraud [56,62,63]. Secondly, different forms of ES offer varying degrees of flexibility [21]. GES allows enterprises to determine their optimal approach to green innovation, while TES provides more specific direction and guidance [21,64]. Thirdly, the content of environmental subsidies differently impacts enterprise operations. PES are more directly integrated with production processes compared to EES. Companies receiving PES typically need to implement comprehensive changes across their production systems, requiring greater organizational effort and commitment. Heavy-polluting enterprises may adopt different strategies based on the characteristics of received subsidies. For instance, when faced with EES, TES, or PES, enterprises often experience increased pressure, potentially leading them to prioritize short-term compliance over long-term GI strategies. Based on these observations, this paper presents the following hypotheses.

H2. 

ES with different characteristics has different impacts on GI.

2.3. Heterogeneity of Property Rights, Environmental Investments, and Earnings Management

From the perspective of ES allocation, several factors influence the effectiveness of ES on GI, including enterprise property rights, environmental investment, and earnings management practices. In China, enterprises are categorized into state-owned and non-state-owned entities [65], with state-owned enterprises typically demonstrating greater commitment to social performance than their non-state-owned counterparts [27,44,66]. Environmental investment, defined as direct expenditure on environmental protection to reduce pollution and enhance resource utilization efficiency [67], plays a crucial role. ES can effectively motivate enterprises with environmental investments to implement GI measures and enhance their innovation capabilities [68]. Corporate earnings manipulation also impacts the relationship between ES and GI. This practice involves deliberately manipulating financial statements to reduce costs and increase profits for private benefits [69]. Due to information asymmetry between governmental authorities and enterprises, companies may secure ES through earnings manipulation, potentially disrupting the optimal allocation of environmental subsidies among enterprises [70,71].

The effects of ES on GI vary depending on enterprises’ property rights structures, environmental investments, and earnings management practices. First, state-owned enterprises, due to their close governmental ties, bear greater social responsibilities and tend to favor traditional projects over GI [61]. In contrast, non-state-owned enterprises demonstrate a stronger inclination to expand their market presence through GI [27]. Second, the high-risk nature of GI makes it particularly sensitive to corporate environmental investments [67]. Drawing on data from the China Ecological Environment Statistical Bulletin 2022, it is evident that the environmental governance expenditure for the completion and acceptance of industrial pollution source treatment and construction projects constituted 33.7% of the total expenditure. Social environmental governance expenditure primarily took the form of investment in urban environmental infrastructure construction. Notably, the industrial sector’s relatively low inclination to invest in environmental protection combined with high transformation costs, may diminish the effectiveness of ES. Some enterprises even view ES primarily as an additional revenue stream. Finally, some enterprises employ earnings management strategies to secure more ES, which undermines GI efforts by companies genuinely committed to environmental sustainability [31]. Moreover, enterprises engaging in earnings management typically prioritize short-term financial objectives [71].

H3. 

Impacts of ES on GI are different for heavy polluting enterprises with different property rights, environmental investments and earnings management levels.

2.4. Moderating Effect of Environmental Enforcement

EE is a process through which public agents detect and sanction violations of environmental protection laws [36,72]. As a typical environmental regulation, EE is considered an effective command-control policy tool [30,73]. According to the deterrence mechanism, enterprises subject to EE are likely to cease their environmental violations and adopt GI activities to maintain organizational legitimacy [74].

Scholars hold divergent views regarding the relationship between EE and GI in enterprises. Some researchers argue that EE can effectively promote GI [73]. For instance, Testa et al. [30] examined the relationship between environmental regulation and firms’ competitive performance, finding that frequent government inspections can stimulate enterprise technology adoption and product innovation. However, other studies suggest that EE may not effectively promote GI. Prechel and Zheng [75], analyzing data from Standard & Poor’s 500 corporations, found that government penalties did not motivate companies to invest in pollution reduction technologies. Furthermore, Liao [24] compared the effectiveness of various environmental policy tools on enterprises’ GI and concluded that strict penalties have limited impact on innovation compared to market measures and information-based instruments.

As a crucial incentive for promoting environmental protection among firms, EE significantly influences the effectiveness of ES for GI [76]. First, non-compliant enterprises often face severe penalties that can result in costs far exceeding the initial investment required for implementing GI [35]. Consequently, heavy-polluting enterprises tend to allocate more ES towards GI activities. Second, EE can lead to market-based penalties, particularly affecting corporate reputation [77]. This motivates enterprises to utilize ES for GI to ensure compliance with regulations. Finally, EE enhances the effectiveness of ES by raising environmental awareness among enterprises [34,78]. These factors demonstrate the complementary relationship between ES and EE in promoting GI. Based on these observations, this paper proposes the following hypothesis.

Meanwhile, the government’s stringent EE can mitigate the inefficiencies in ES caused by property rights, environmental investments, and earnings management. Under environmental penalty policies, state-owned enterprises are typically held to higher standards, compelling them to actively pursue GI to prevent negative environmental impacts [62,79,80]. Furthermore, enhanced EE creates external pressure on heavy-polluting enterprises, forcing those with limited environmental investments to innovate and comply with regulatory requirements to avoid penalties [35]. Additionally, EE weakens the ability of heavy-polluting enterprises to manipulate earnings, limiting managers to earnings management through actual activities, which ultimately facilitates better implementation of GI initiatives [81]. Based on these observations, this article proposes the following hypothesis.

H4. 

EE can play a positive moderating role in the process of ES influencing GI.

3. Search Design

3.1. Sample and Data Resource

The sample for this study consists of enterprises from heavy-polluting industries listed in China’s A-share market. Based on the Environment Information Disclosure Guidance for Listed Companies issued by China’s Ministry of Environmental Protection in 2010 and the revised Guideline for the Industry Classification of Listed Companies by the Communication Science Research Center in 2012, sixteen categories of heavy-polluting industries were identified, including encompass thermal power, iron and steel, coal, metallurgy, chemicals, petrochemicals, building materials, paper making, pharmaceuticals, textiles, leather-making, and mining. These industries correspond to the 2-digit industry codes B06-10, C17, C19, C22, C25-32, and D44.

The study period spans from 2012 to 2021 to ensure sample selection consistency. Companies with ST, *ST, and PT status were excluded from the analysis. Additionally, firms with incomplete data or significant outliers were removed, specifically those with asset-liability ratios exceeding 1 or negative returns on total assets. After applying these screening criteria, the final sample comprises 1254 A-share companies from heavy-polluting industries, yielding a total of 8422 observations.

Data on environmental subsidies (ES) are collected through systematic manual analysis of annual reports from heavily polluting enterprises. To ensure data accuracy and reliability, we implemented a three-step collection process. First, we identified the scope of subsidies based on the primary types of ES received by enterprises, which encompass pollution monitoring subsidies, energy conservation subsidies, emission reduction subsidies, pollution abatement subsidies, other environmental subsidies, and environmental protection awards. The detailed definitions of these ES categories are presented in

Table 1. ES types.

ES Type	Content
Pollution monitoring subsidy	Subsidy for pollution prevention equipment and technology.
Energy conservation subsidy	Subsidy for activities including conservation in energy use and the recycling and comprehensive utilization of pollutants and secondary energy sources (waste heat, waste gas, etc.).
Emission reduction subsidy	Subsidy for activities including the use of clean energy, desulfurization, denitration, emission reduction in volatile organic compounds, elimination of vehicles with high-pollution emissions.
Pollution abatement subsidy	Subsidy for the treatment of environmental pollution caused by pollutants (waste solid, water, and gas).
Subsidy with no specific targets	Subsidy that does not specify the purpose of the project (e.g., funds for environmental protection projects and subsidies for green economic development).
Environmental protection awards	Financial awards granted to enterprises or individuals for their outstanding contributions to environmental protection.

. Second, we divided our research team into two independent groups to analyze subsidy information in the annual reports and document the annual subsidy amounts received by each enterprise across all ES categories. Finally, we cross-validated the findings between the two groups to determine the definitive ES figures for each enterprise.

This article utilizes multiple data sources. The GI data is obtained from the Chinese Research Data Services Platform, which aggregates patents from both the China National Intellectual Property Administration and Google Patent. These patents are categorized according to the green patent criteria established by the World Intellectual Property Office [82,83]. EE data comes from the Institute of Public Environment Affairs (IPE), a nonprofit environmental research institute established in 2006. The IPE maintains a comprehensive enterprise environmental performance database that documents government environmental supervision, including detailed information about enterprises, violation specifics, and corresponding penalties. Additional enterprise-related data is extracted from the Wind database.

3.2. Variables

(1) The explained variable is GI. Due to its inherent complexity, measuring and quantifying GI presents significant challenges. While various metrics have been developed to assess GI, patents remain the most widely used indicator [38]. Although patents serve as intermediate measures, they effectively demonstrate the firms’ capabilities and commitment to scientific advancement in green development [62]. Patent-related measurements of GI encompass patent applications, granted patents, and more comprehensive metrics, such as patent ratios [84,85]. Green patent applications, however, typically undergo less rigorous review processes and may not fully validate the innovativeness or environmental benefits of the proposed technology. While patent ratios, which calculate the proportion of green patents, can indicate a company’s green development level, this metric is vulnerable to fluctuations in the firm’s overall innovation performance. Authorized green patents provide a direct measure of concrete scientific and technological innovation outcomes [85]. These patents undergo thorough examination, making them a more reliable indicator. The number of authorized green patents a company holds directly reflects its investment in green technology innovation and demonstrates its GI capabilities [86]. Therefore, this study employs the logarithm of authorized green patents as the measurement for GI.

(2) The explanatory variable ES represents government subsidies provided to enterprises for their environmental protection initiatives. We collected data on environment-related subsidies recorded in the non-operating income section of companies’ annual reports. The various categories of environmental subsidies received by companies are listed in Table 1. We summed these subsidies and took the logarithm to derive our core explanatory variable.

This paper further categorizes ES based on their timing, flexibility, and content characteristics. First, environmental subsidies are classified as ESB and ESA, depending on when they are granted. Second, subsidies are categorized as GES or TES based on whether they target specific environmental projects. Finally, considering the two types of environmental protection behaviors in relation to production activities, subsidies are divided into PES and EES. The specific characteristics of these environmental subsidies are detailed below.

• ESB: Government funding provided directly to companies that implement environmentally protective measures.
• ESA: Government awards granted to companies for implementing environmentally protective practices and initiatives.
• TES: ES that define the project’s purpose and objectives.
• GES: ES that does not specify a particular project purpose or scope.
• PES: Subsidies provided for pollution prevention activities integrated into production processes.
• EES: Subsidies specifically designed for end-of-pipe pollution control measures that operate independently of the main production processes.

By analyzing various types of environmental subsidies (ES) shown in Table 1 and their inherent characteristics, we have derived the subsidy characteristic variables.

Table 2. ES and ES characteristics.

ES Types	ES Characteristics
Classification	Subsidy Timing		Subsidy Form		Subsidy Content
	ESB	ESA	TES	GES	PES	EES
Pollution monitoring subsidy	√		√		√
Energy conservation subsidy	√		√		√
Emission reduction subsidy	√		√		√
Pollution abatement subsidy	√		√			√
Subsidy with no specific targets	√			√
Environmental protection awards		√

Note: This table shows the results after reclassifying the different forms of ES by time, form, and content. “√” explains the type of each subsidy in the first column according to time, form and content. The backgraound color means that the corresponding ES are not considered in the classification process.

categorizes five distinct types of ES based on their defining features. It should be noted that, due to limited disclosure of detailed subsidy information, environmental protection rewards have been excluded from our classification of subsidy form and content. Additionally, subsidies without specific targets have been omitted from the subsidy content classification.

There is a key point about the scope difference between ESB and ESA. The existing literature primarily distinguishes between ESB and ESA using two methods. The first method considers ESB as environmental subsidies in deferred revenue, while ESA are the ones in the current profit and loss [56]. The second approach distinguishes between tax-based and non-tax-based environmental subsidies [87,88]. However, both approaches have limitations. Under Chinese accounting standards, certain subsidies within current profit and loss can be classified as ESB, contradicting the first approach. Additionally, environmental tax-based subsidies are rarely disclosed separately from non-environmental tax-based subsidies, limiting the effectiveness of the second approach. In this paper, we define ESB as direct government funding intended to encourage future environmental protection initiatives, while ESA represents rewards for existing environmental protection achievements. Specifically, ESB is typically allocated before the completion of environmental protection projects, whereas ESA is disbursed upon project completion or near completion.

(3) Moderating variable: EE. EE has a moderating effect mainly through the compliance pressure it exerts on heavy-polluting enterprises [4,89,90]. In this study, EE is measured by examining whether companies have been subject to environmental penalties, administrative orders, or administrative ratings during the sample period and previous years. According to the Measures for Environmental Administrative Punishment (Ministry of Environment Protection, 2010), administrative penalties encompass fines, warnings, administrative detention, production restrictions, and facility shutdowns. Administrative orders require companies to cease polluting activities, implement corrective measures, or modify production processes to prevent environmental pollution. Administrative ratings identify companies that receive lower-tier classifications in government environmental performance evaluations. This includes enterprises issued yellow cards, red cards, or black cards due to hazardous operations. Additionally, it covers companies designated as C-level and D-level enterprises, which are identified as key concerns for emergency emission reductions during periods of severe pollution. And the situations associated with significant risks of sudden environmental are included.

(4) Control variables in this study primarily focus on enterprise characteristics [11,49]. The controlled variables include R&D intensity (RD), company size (Size), company age (Age), financial leverage (Lev), nature of equity (Soe), cash flow ratio (Cash), and return on assets (Roa). These variables significantly influence firms’ green innovation capabilities [27]. Size is measured by the natural logarithm of total assets, reflecting the advantages that large enterprises possess in resources and innovation activities. Age is calculated using the logarithm of the enterprise establishment period with the natural base. The relationship between age and innovation capability is complex: while accumulated experience and knowledge can enhance innovation capability over time, organizational inertia may impede innovation. Lev is determined by the ratio of liabilities to total assets, indicating a firm’s leverage level and financial risk. Soe differentiates between state-owned enterprises (SOE) and non-SOE, acknowledging that companies with different ownership structures demonstrate varying behaviors in GI. Cash measures a company’s liquidity and investment capacity for GI through the ratio of operating and investing activity cash to total assets. Roa is the ratio of net profit income after tax to total assets, evaluating corporate performance. Strong operational performance enables reinvestment and innovation activities. RD is represented by the natural logarithm of one plus R&D expenditure per employee, indicating the company’s research and development commitment.

This study also controls industry dummy variables. Heavy-polluting industries exhibit significant differences in pollution control difficulty and environmental protection technology levels, resulting in distinct challenges in green transformation across industries. This variation likely leads to substantial differences in ES application probabilities among enterprises in different heavy-polluting industries. Furthermore, government support may vary by industry, necessitating the control of heavy-polluting industry dummy variables. Regional factors are also significant. Heavy-polluting enterprises in different provincial regions face varying ecological environmental carrying capacities and local government environmental protection efforts, affecting their capabilities and probabilities of obtaining ES. Year fixed effects are equally important, as they control for the impact of macro-environmental policy adjustments and changes in environmental protection conditions on heavy-polluting enterprises’ motivation to apply for ES [91]. The variables and their measurements are shown in

Table 3. Variables and measurements.

Type	Variable	Symbols	Measurement
Explanatory Variable	Environmental Subsidies	ES	Natural logarithm of (1 + environmental-related subsidies)
	ES characteristic	ESB	Natural logarithm of (1 + ES beforehand in Table 1)
		ESA	Natural logarithm of (1 + ES afterwards in Table 1)
		TES	Natural logarithm of (1 + targeted ES in Table 1)
		GES	Natural logarithm of (1 + general ES in Table 1)
		PES	Natural logarithm of (1 + pollution prevention ES in Table 1)
		EES	Natural logarithm of (1 + end-of-pipe ES in Table 1)
Explained Variable	Green innovation	GI	Natural logarithm of 1 plus the application number of green patent
Moderating Variable	Environmental enforcement	EE	1 for the year that the firm is enforced and all subsequent years, otherwise, 0 (dummy)
Control Variable	Financial leverage	Lev	Liabilities/total assets
	R&D intensity	RD	Natural logarithm of 1 plus the R&D expenditures per employee
	Company size	Size	Natural logarithms of total assets
	Cash flow ratio	Cash	Operating activity cash and investing activity cash/total assets
	Return on assets	Roa	Net profit income after tax/total assets
	Company age	Age	Logarithm of the enterprise establishment period using the natural base
	Natural of equity	Soe	1 for the state-owned firm, otherwise, 0
	Year	Year	Annual control dummy variable
	Province	Prov	Province control dummy variable
	Industry	Ind	Industry control dummy variable

.

3.3. Model

To examine the relationships among ES, ES characteristics, and GI, as well as to analyze the moderating effects of EE, this study employs a four-model framework. First, Model (1) investigates the direct impact of ES on GI. Second, Model (2) analyzes how different ES characteristics influence GI through separate regression analyses, focusing on three aspects: time-point features (before and after subsidy, ESB and ESA), flexibility degrees (targeted and general subsidy, TES and GES), and content features (production prevention subsidy and end-of-pipe control subsidy, PES and EES). Finally, Models (3) and (4) assess the moderating effect of EE on these relationships.

$${\mathrm{G}\mathrm{I}}_{\mathrm{i}\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{E}\mathrm{S}}_{\mathrm{i}\mathrm{t}-1}+{\mathsf{\alpha }}_2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\alpha }}_3\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}+{\mathsf{\alpha }}_4\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}+{\mathsf{\alpha }}_5\mathrm{I}\mathrm{n}\mathrm{d}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(1)

$${\mathrm{G}\mathrm{I}}_{\mathrm{i}\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{E}\mathrm{S} \mathrm{C}\mathrm{h}\mathrm{a}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{s}}_{\mathrm{i}\mathrm{t}-1}+{\mathsf{\beta }}_3{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\beta }}_4\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}+{\mathsf{\beta }}_5\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}+{\mathsf{\beta }}_6\mathrm{I}\mathrm{n}\mathrm{d}+{\mathsf{\delta }}_{\mathrm{i}\mathrm{t}}$$

(2)

$${\mathrm{G}\mathrm{I}}_{\mathrm{i}\mathrm{t}}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_1{\mathrm{E}\mathrm{S}}_{\mathrm{i}\mathrm{t}-1}+{\mathsf{\gamma }}_2{\mathrm{E}\mathrm{E}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\gamma }}_3{\mathrm{E}\mathrm{S}}_{\mathrm{i}\mathrm{t}-1}\#{\mathrm{E}\mathrm{E}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\gamma }}_4{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\gamma }}_5\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}+{\mathsf{\gamma }}_6\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}+{\mathsf{\gamma }}_7\mathrm{I}\mathrm{n}\mathrm{d}+{\mathsf{ϵ}}_{\mathrm{i}\mathrm{t}}$$

(3)

$${\mathrm{G}\mathrm{I}}_{\mathrm{i}\mathrm{t}}={\mathsf{\delta }}_0+{\mathsf{\delta }}_1{\mathrm{E}\mathrm{S} \mathrm{C}\mathrm{h}\mathrm{a}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{s}}_{\mathrm{i}\mathrm{t}-1}+{\mathsf{\delta }}_2{\mathrm{E}\mathrm{E}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\delta }}_3{\mathrm{E}\mathrm{S} \mathrm{C}\mathrm{h}\mathrm{a}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{s}}_{\mathrm{i}\mathrm{t}-1}\#{\mathrm{E}\mathrm{E}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\delta }}_4{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\delta }}_5\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}+{\mathsf{\delta }}_6\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}+{\mathsf{\delta }}_7\mathrm{I}\mathrm{n}\mathrm{d}+{\mathsf{\vartheta }}_{\mathrm{i}\mathrm{t}}$$

(4)

For the above models, “i” represents enterprises, “t” indicates the year, and the symbol “#” represents interaction terms. “Controls” are those variables that may have impacts on GI. Additionally, this study controls for temporal effects through the “Year” variable, industry-specific characteristics (denoted as ‘Ind’), and regional factors (denoted as ‘Prov’).

To address potential endogeneity concerns in our research, we implemented two strategic approaches. First, we incorporated a one-period lag for ES and their various characteristics, which effectively mitigates reverse causality and spurious regression issues [92]. Second, we addressed the sample selection bias, where firms with stronger GI capabilities tend to be more proactive in seeking and securing ES. To resolve this, we utilized an instrumental variable approach, employing the average subsidy allocation at the 2-digit industry code level for ES and their characteristics. The higher the average level of industrial ES, the greater the probability that an enterprise will obtain environment-related subsidies, but it has no direct relationship with the enterprise’s GI [93]. Further, we employ the Two-Stage Least Squares (2SLS) estimation method to estimate the above models.

4. Results

4.1. Descriptive Statistics

The distribution of heavy-polluting enterprises in China across industries and time from 2012 to 2021 is shown in

Table 4. Statistics of Listed Companies in Heavy-polluting Industries (2012–2021).

	Thermal Power	Iron and Steel	Coal	Metallurgy	Building Materials	Mining	Chemicals	Petrochemicals	Pharmaceuticals	Papermaking	Textiles	Leather-Making
2012	57	28	21	42	65	27	196	16	126	20	58	6
2013	58	29	24	46	64	30	205	16	128	24	62	6
2014	59	28	24	51	74	30	219	17	138	23	56	6
2015	61	29	25	55	77	32	236	15	153	23	58	8
2016	62	29	23	56	78	30	259	15	166	28	59	8
2017	67	26	22	66	81	30	288	17	195	28	59	11
2018	63	28	23	69	78	30	290	16	201	29	61	11
2019	63	29	22	63	79	31	294	15	209	27	54	10
2020	66	29	22	69	92	33	327	16	233	32	60	11
2021	74	30	22	76	97	32	362	17	257	35	70	11

. The data reveals significant variations in enterprise numbers across different industrial sectors. The chemical and pharmaceutical industries demonstrate the highest concentration of enterprises, while the leather-making industry shows the lowest representation. Over this period, substantial growth in enterprise numbers has been observed in the chemical, pharmaceutical, metallurgical, and building material sectors. In contrast, industries such as thermal power, iron and steel, coal, mining, and petrochemicals have maintained relatively stable numbers. This trend may indicate that many heavy-polluting industries have begun transitioning toward green transformation.

Figure 1 illustrates the annual mean values of ES and EE. The average ES shows a declining trend since 2019, while the average EE consistently increases, particularly since 2015. These trends suggest a period of adjustment for ES and a growing stringency in EE.

The descriptive statistics for all variables are presented in

Table 5. Variable Descriptive Statistics (N = 8422).

Variables	Mean	Std. Dev.	10% Quantile	50% Quantile	90% Quantile
GI	3.66	25.23	0.00	0.00	6.00
ES	7.72	6.94	0.00	11.00	15.61
ESB	7.35	6.93	0.00	10.49	15.42
ESA	1.59	4.31	0.00	0.00	10.82
GES	2.61	5.27	0.00	0.00	12.78
TES	6.63	6.86	0.00	0.00	15.08
PES	5.95	6.75	0.00	0.00	14.81
EES	2.54	5.16	0.00	0.00	12.76
EE	0.27	0.44	0.00	0.00	1.00
Size	12.74	0.32	12.28	12.74	13.23
Age	2.97	0.29	2.64	3.00	3.30
Lev	45.25	4.33	40.43	44.44	55.82
Roa	0.01	0.01	−0.01	0.01	0.03
Cash	−0.02	0.02	−0.06	−0.02	0.00
Soe	0.36	0.48	0.00	0.00	1.00
RD	1.26	0.82	0.05	1.30	2.31

. The data reveals that more than 50% of heavy-polluting enterprises filed no green patents during the study period, indicating a significant lack of enthusiasm for GI activities. Analysis of the mean values of ES across different categories shows that ESB is higher than ESA. Similarly, TES exceed general GES, while PES are greater than EES. These findings indicate that the majority of ES are disbursed before the completion of environmental projects, primarily in the form of targeted funding for pollution prevention initiatives. The median value (50th percentile) of EE remains at zero, suggesting that only a small proportion of enterprises have faced governmental enforcement actions during the study period.

Table 6. Correlations.

	GI	ES	ESB	ESA	GES	TES	PES	EES	EE
GI	1
ES	0.2152 *	1
ESB	0.2219 *	0.9597 *	1
ESA	0.0750 *	0.3697 *	0.2287 *	1
GES	0.1857 *	0.4923 *	0.5111 *	0.1731 *	1
TES	0.1866 *	0.8798 *	0.9159 *	0.2421 *	0.2984 *	1
PES	0.1906 *	0.8136 *	0.8462 *	0.2535 *	0.2872 *	0.9241 *	1
EES	0.1232 *	0.4812 *	0.5003 *	0.1708 *	0.2567 *	0.5436 *	0.3522 *	1
EE	0.1501 *	0.1312 *	0.1351 *	0.0327 *	0.0966 *	0.1274 *	0.1182 *	0.1129 *	1

Note: The symbol * denotes significance at the 10% levels.

presents the correlations between various variables, including ES, ES characteristics, EE, and GI. All these variables show significant correlations significantly. This means that that environmental strategies, such as ES and EE, can affect GI ability of heavy-polluting firms significantly.

4.2. Empirical Results

4.2.1. Effects of ES on GI

We conducted regression analyses and computed variance inflation factors (VIFs) to evaluate potential multicollinearity. The results revealed that all VIFs fell well below the critical threshold of 10, indicating no significant multicollinearity concerns. Additionally, we performed a weak instrument test using the Cragg-Donald Wald F statistic. The analysis showed that all F values surpassed 10, confirming the robustness and validity of our chosen instrumental variables.

Table 7. The impact of ES on GI.

	Explained Variable: GI
ES		−0.0625 **	−0.0583 **
		(−2.1802)	(−2.1733)
Size	0.4285 ***	0.3550	0.3608
	(8.1299)	(1.0723)	(1.1344)
Age	0.0809 **	0.0897	0.1849 **
	(1.9895)	(1.1398)	(2.1779)
Roa	0.8288	0.1644	0.1604
	(0.7969)	(0.1161)	(0.1180)
Soe	0.3050 ***	0.5264 ***	0.4393 ***
	(12.8685)	(7.0887)	(6.4085)
Lev	0.0053	0.0053	0.0055
	(1.1007)	(0.8548)	(0.9299)
Cash	0.2645	0.5831	0.5613
	(0.2158)	(0.2171)	(0.2175)
RD	0.1797 ***	0.1826 ***	0.1669 ***
	(12.1609)	(7.2986)	(7.4962)
_cons	−5.3141 ***	−3.6311	−4.3303
	(−6.5913)	(−0.8205)	(−1.0224)
Year dummy	Yes	Yes	Yes
Industry dummy	Yes	Yes	Yes
Province dummy	Yes	No	Yes
Cragg-Donalf Wald F		124.543	117.33
N	8298	6998	6998

Note: The symbols *** and ** denote significance at the 1% and 5% levels, respectively.

demonstrates the impact of ES on GI among heavy-polluting firms. Due to substantial variations in characteristics and development patterns across heavy-polluting industries, the second column of regression results incorporates both year and industry fixed effects. Additionally, the third column accounts for provincial control variables, considering regional differences in industrial structure, economic development levels, and industrial policies.

The regression analysis reveals a significant negative correlation between ES and GI, indicating that heavy-polluting companies receiving higher ES are less likely to engage in GI activities. This finding supports Hypothesis H1, which demonstrates the inefficiency of ES in influencing GI for heavy-polluting enterprises in China. The negative impact of ES on GI has also been identified in some previous studies [18,30,31,47,49,50]. This negative outcome may indicate the following facts. First, the intensity of subsidies remains insufficient [49]. The relatively low levels of ES provided to heavy-polluting enterprises in China are insufficient to offset the costs and externalities associated with GI. Second, there may be issues in the design of ES. Only “well-designed” subsidies can better exert their effectiveness [30,47]. Third, there are problems in the allocation of ES. The misallocation of subsidy funds among enterprises weakens the effect of subsidies on GI [31,49,50]. The negative effect of ES on GI is exactly the opposite of the results of Shao and Chen [27] and Wei et al. [28]. This is because the above-mentioned studies analyzed all enterprises in China. This opposite result also demonstrates the particularity and importance of the role of ES heavy-polluting enterprises’ GI.

Table 8. The impacts of ES characteristics on GI.

	Explained Variable: GI
	ES Timing	ES Form	ES Content
ESB	−0.0527 ***
	(−3.7016)
ESA	0.0161
	(0.6467)
TES		−0.0681 ***
		(−4.0695)
GES		0.0679 ***
		(3.3689)
PES			−0.0489 ***
			(−2.9927)
EES			0.0402 **
			(2.4649)

Note: The symbols *** and ** denote significance at the 1% and 5% levels, respectively.

illustrates the differential impacts of ES with varying characteristics on GI in heavy-polluting enterprises. The analysis focuses on key explanatory variables, with control variable results excluded for clarity. The impact of ES varies significantly across different characteristics. Regarding timing, ESB demonstrates a significant negative impact on GI, while ESA shows no significant effect. In terms of subsidy forms, TES exhibits a significant negative influence, whereas GES shows a significant positive impact. Similarly, for subsidy content, PES has a significant negative effect, while EES demonstrates a significant positive influence. These findings align with Hypothesis 2. When heavy-polluting enterprises receive ESB, TES, or PES, they must meet stringent acceptance criteria, requiring substantial financial allocation to achieve specific targets. Consequently, the heightened environmental pressure forces these enterprises to prioritize immediate pollution control over long-term GI investments.

To analyze the heterogeneous effects of ES on GI across different enterprises, this paper categorizes enterprises based on three criteria: property rights, environmental protection investment status, and earnings management practices. Heterogeneity analysis can reveal the problems of ES allocation. The empirical tests are conducted using models (1) and (3) with the following classification methodology: First, enterprises are categorized into state-owned enterprises (SOE) and non-state-owned enterprises (non-SOE). Second, they are classified as enterprises with environmental protection investment (EPI) and those without (non-EPI). Following Zhu and Mao’s [94] research methodology, environmental protection investment data is manually collected from enterprise annual reports. This includes environmental assessment fees, sewage discharge fees, and greening fees listed under “administrative expenses” in the income statement. Additionally, it encompasses expenditures for desulfurization and denitrification projects, sewage and waste gas treatment facilities, waste heat power generation, energy conservation initiatives, clean production lines, and dust removal systems detailed in the “construction in progress” section. Finally, enterprises are divided into high earnings management (REM) and low earnings management (non-REM) categories based on their earnings management levels. Following Roychowdhury’s [95] methodology, three types of real earnings management are measured using least squares regression while controlling for year and industry variables, sales manipulation, production manipulation, and discretionary expense manipulation. A comprehensive real earnings management index is constructed using Cohen [96] approach. Subsequently, a dummy variable REM is established to measure enterprise earnings management. Enterprises with real earnings management levels above the sample median are classified as REM enterprises, while those below are designated as non-REM enterprises.

Then, we analyzed the impact of ES on GI across different groups (

Table 9. Heterogeneity of the effects of ES on GI.

	Explained Variable: GI
	SOE	Non-SOE	EPI	Non-EPI	REM	Non-REM
ES	−0.1243 ***	0.0536 ***	0.0191	−0.1395 ***	−0.1025 ***	−0.0056
	(−3.8713)	(3.2428)	(0.8483)	(−3.0539)	(−2.9132)	(−0.2144)
Controls	Yes
Year dummy	Yes
Industry dummy	Yes
Province dummy	Yes
N	2563	4435	3050	3948	3547	3451

Note: The symbol *** denotes significance at the 1% levels.

). The finding supports Hypothesis H3. The results reveal that ES has a significant negative effect on GI in SOE enterprises. The SOE enterprises have lower motivation for GI and poorer utilization of ES, which has been observed in the previous literature [27,66]. For non-EPI enterprises, the negative impact is particularly significant, as these companies lack the fundamental environmental protection infrastructure and technology. Consequently, even when receiving subsidies, they struggle to implement GI projects effectively. Similarly, the effect of ES on GI in enterprises engaging in earnings management shows a significant negative relationship, indicating that such practices may obscure their true operational conditions and financial risks. These companies tend to manipulate profits to achieve short-term performance targets while neglecting investments in long-term green development. As Li and Xiao [31] and Bai et al. [49] stressed, enterprises spent more time and effort on obtaining subsidies rather than innovation, which weaken the effectiveness of the policy.

4.2.2. Moderating Effects of EE

Table 10. The moderating effects of EE.

	Explained Variable: GI
	Overall Effect	ES Timing	ES Form	ES Content
ES#EE	0.1213 ***
	(5.9909)
ESB#EE		0.1148 ***
		(5.5154)
ESA#EE		−0.0076
		(−0.2265)
TES#EE			0.1039 ***
			(4.7071)
GES#EE			−0.0178
			(−0.6419)
PES#EE				0.0866 ***
				(4.1659)
EES#EE				−0.0168
				(−0.7397)

Note: The symbol *** denotes significance at the 1% levels. Symbol “#” means variable interaction.

presents the results of the moderating effect of EE corresponding to models (3) and (4). In the process where ES influences GI, EE demonstrates a significantly positive moderating effect, confirming Hypothesis H4. This indicates that stringent EE not only mitigates potential adverse impacts of ES, but also substantially enhances its positive influence on GI. When examining the moderating effect of EE on subsidy effectiveness across different characteristics, the regression coefficients for the interaction terms between ESB, TES, and PES are all significantly positive, validating Hypothesis H4. These findings indicate that various types of ES, when coupled with strict EE, demonstrate more pronounced facilitative effects. This suggests that stringent EE motivates enterprises to prioritize GI, thereby improving the efficiency and effectiveness of subsidy utilization. Additionally, our study reveals that flexible policy frameworks are more effective at stimulating GI compared to targeted policies. A flexible policy environment enables enterprises to explore diverse GI pathways aligned with their specific characteristics while maintaining basic environmental standards.

Table 11. Heterogeneity of the moderating effects of EE.

	Explained Variable: GI
	SOE	Non-SOE	EPI	Non-EPI	REM	Non-REM
ES	−0.1664 ***	0.0631 ***	0.0074	−0.1813 ***	−0.1628 ***	−0.0159
	(−4.1874)	(2.6186)	(0.2536)	(−3.1156)	(−3.3794)	(−0.5010)
ES#EE	0.1000 ***	−0.0187	0.0202	0.1248 ***	0.1070 ***	0.0306 *
	(5.4240)	(−1.3045)	(1.5411)	(3.5397)	(4.3600)	(1.7112)
Controls	Yes
Year dummy	Yes
Industry dummy	Yes
Province dummy	Yes
N	2563	4435	3050	3948	3547	3451

Note: The symbols *** and * denote significance at the 1% and 10% levels, respectively. Symbol “#” means variable interaction.

presents the interaction effects between ES and EE on GI in heavily polluting enterprises. The results reveal that the interaction coefficient is significantly positive for SOE enterprises, non-EPI enterprises and REM enterprises. This finding suggests that stringent environmental penalties compel heavily polluting enterprises to pursue GI, thereby enhancing the effectiveness of ES. Consequently, EE serves as an effective mechanism to address the inefficient allocation of ES. These results in Table 8 and Table 9 align with the conclusions drawn by some studies focusing on the green transition issues in developing countries [21,28,97,98]. EE is widely recognized as one of the most effective drivers of GI. Therefore, relying solely on subsidies to promote corporate GI is insufficient [30]; it must be complemented by stringent EE to create a robust policy package that can more effectively stimulate corporate GI activities [24].

4.3. Robustness Tests

4.3.1. Heckman Two-Step Method

When analyzing the relationship between GI and ES, there exists a potential sample selection bias. Firms may receive subsidies based on specific characteristics, such as size, age, or state ownership status, which could simultaneously influence their patent authorization behavior. If not properly addressed, this bias could lead to inaccurate estimates. To mitigate this issue, we employ the Heckman two-step method [99]. This approach consists of two distinct stages. First, a Probit model predicts firm selection into the main regression model by estimating the probability of receiving subsidies based on observable firm characteristics. Second, the derived Mills ratio (or inverse Mills ratio) is incorporated as an additional explanatory variable in the main regression equation to control for sample selection bias. The Heckman correction effectively addresses estimation biases arising from sample selection issues, resulting in more accurate and reliable regression results. By accounting for the non-random selection process in subsidy allocation, this method ensures that the estimated relationships between GI and ES reflect genuine causal effects rather than spurious correlations driven by sample selection. We re-examined the relationship between ES and ES with different characteristics and GI using the Heckman two-step method, and the results are robust. The results are shown in

Table 12. Heckman two-step method Robustness tests.

	Explained Variable: GI
ES	−0.0583 ***
	(−4.1378)
ESA		0.0161
		(0.6467)
ESB		−0.0527 ***
		(−3.7016)
TES			−0.0681 ***
			(−4.0695)
GES			0.0679 ***
			(3.3689)
PES				−0.0489 ***
				(−2.9927)
EES				0.0402 **
				(2.4649)
Year dummy	Yes	Yes	Yes	Yes
Industry dummy	Yes	Yes	Yes	Yes
Province dummy	Yes	Yes	Yes	Yes
Controls	Yes	Yes	Yes	Yes
N	6998.0000	6998.0000	6998.0000	6998.0000

Note: The symbols *** and ** denote significance at the 1% and 5% levels, respectively.

.

4.3.2. Propensity Score Match

To validate our research findings and mitigate potential sample issues derived from EE, we implemented a robustness check using the Propensity Score Matching (PSM) method [84,85]. Since EE is not randomly distributed among firms, potential heterogeneity concerns needed to be addressed. We systematically matched firms with EE to comparable firms without EE, eliminating unmatched observations, and subsequently reassessed EE’s moderating effects. The validation process incorporated three distinct matching techniques, 1-to-1 nearest neighbor matching, radius caliper matching, and kernel matching. Across all three matching approaches, our primary findings remained robust and consistent, thereby substantiating the reliability and validity of our research conclusions. To simplify the presentation of the results, we have chosen to report only the findings from kernel matching in the main text (

Table 13. Kernel matching robustness tests.

	Explained Variable: GI
ES	−0.1179 ***
	(−4.7875)
ES#EE	0.1389 ***
	(6.0431)
ESB		−0.1118 ***
		(−4.3948)
ESA		0.0149
		(0.4028)
ESB#EE		0.1337 ***
		(5.5291)
ESA#EE		−0.0130
		(−0.3482)
TES			−0.1170 ***
			(−4.4633)
GES			0.0551 *
			(1.7796)
TES#EE			0.1245 ***
			(4.9261)
GES#EE			−0.0252
			(−0.8318)
PES				−0.0901 ***
				(−3.6939)
EES				0.0308
				(1.1968)
PES#EE				0.1052 ***
				(4.4823)
EES#EE				−0.0216
				(−0.8543)
EE	−0.8814 ***	−0.7787 ***	−0.5929 ***	−0.4010 ***
	(−5.0898)	(−4.6928)	(−3.7758)	(−2.8539)
Year dummy	Yes	Yes	Yes	Yes
Industry dummy	Yes	Yes	Yes	Yes
Province dummy	Yes	Yes	Yes	Yes
Controls	Yes	Yes	Yes	Yes
N	6928.0000	6928.0000	6928.0000	6928.0000

Note: The symbols *** and * denote significance at the 1% and 10% levels, respectively. Symbol “#” means variable interaction.

). It is noteworthy that the results produced by 1:1 matching and radius caliper matching are qualitatively identical to those obtained through kernel matching, with all relevant coefficients maintaining the same signs.

4.3.3. Alternative Explained Variable

As a robustness check, we employed the natural logarithm of green patent applications as an alternative dependent variable. The results remained consistent with our core findings, further confirming the reliability and stability of our research conclusions. The results are presented in

Table 14. Alternative explained variable for effects of ES and ES characteristics on GI.

	Explained Variable: GI (Green Patent Applications)
ES	−0.0344 ***
	(−2.6175)
ESA		0.0115
		(0.4898)
ESB		−0.0285 **
		(−2.1439)
TES			−0.0572 ***
			(−3.4689)
GES			0.0842 ***
			(4.2167)
PES				−0.0493 ***
				(−2.9749)
EES				0.0655 ***
				(3.9351)
Controls	Yes	Yes	Yes	Yes
Year dummy	Yes	Yes	Yes	Yes
Industry dummy	Yes	Yes	Yes	Yes
Province dummy	Yes	Yes	Yes	Yes
N	7056.0000	7056.0000	7056.0000	7056.0000

Note: The symbols *** and ** denote significance at the 1% and 5% levels, respectively.

and

Table 15. Alternative explained variable for the moderating effects of EE.

	Explained Variable: Green Patent Applications
ES	−0.0652 ***
	(−3.3789)
ES#EE	0.0935 ***
	(5.1795)
ESB		−0.0573 ***
		(−2.9283)
ESA		0.0061
		(0.2008)
ESB#EE		0.0850 ***
		(4.5691)
ESA#EE		−0.0005
		(−0.0166)
TES			−0.0816 ***
			(−3.6947)
GES			0.0769 ***
			(2.7832)
TES#EE			0.0949 ***
			(4.4501)
GES#EE			−0.0445 *
			(−1.6490)
PES				−0.0701 ***
				(−3.2526)
EES				0.0618 ***
				(2.6714)
PES#EE				0.0888 ***
				(4.2846)
EES#EE				−0.0478 **
				(−2.0956)
EE	−0.5408 ***	−0.4487 ***	−0.3510 ***	−0.2376 *
	(−3.9682)	(−3.4493)	(−2.6512)	(−1.8764)
Controls	Yes	Yes	Yes	Yes
Year dummy	Yes	Yes	Yes	Yes
Industry dummy	Yes	Yes	Yes	Yes
Province dummy	Yes	Yes	Yes	Yes
N	7056.0000	7056.0000	7056.0000	7056.0000

Note: The symbols ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. Symbol “#” means variable interaction.

.

4.3.4. Sample Period Adjustment

Due to the 2021 revision of China’s Patent Law, we adjusted our sample period to focus on data from 2012 to 2020. This modification did not affect the validity of our analysis, as our regression results remained robust and consistently supported our initial findings. Results are presented in

Table 16. Sample period adjustment for effects of ES and ES characteristics on GI.

	Explained Variable: GI
ES	−0.0797 ***
	(−4.8637)
ESA		0.0286
		(1.0032)
ESB		−0.0778 ***
		(−4.6185)
TES			−0.0985 ***
			(−4.8926)
GES			0.0905 ***
			(3.7371)
PES				−0.0850 ***
				(−4.1062)
EES				0.0544 ***
				(2.8362)
Year dummy	Yes	Yes	Yes	Yes
Industry dummy	Yes	Yes	Yes	Yes
Province dummy	Yes	Yes	Yes	Yes
Controls	Yes	Yes	Yes	Yes
N	6036.0000	6036.0000	6036.0000	6036.0000

Note: The symbol *** denotes significance at the 1% levels.

and

Table 17. Sample period adjustment for moderating effects of EE.

	Explained Variable: GI
ES	−0.1237 ***
	(−4.9706)
ES#EE	0.1431 ***
	(6.1255)
ESB		−0.1183 ***
		(−4.7923)
ESA		0.0149
		(0.3729)
ESB#EE		0.1385 ***
		(5.8950)
ESA#EE		−0.0130
		(−0.3250)
TES			−0.1296 ***
			(−4.8569)
GES			0.0725 **
			(2.2012)
TES#EE			0.1340 ***
			(5.1743)
GES#EE			−0.0421
			(−1.3108)
PES				−0.1108 ***
				(−4.2029)
EES				0.0369
				(1.3571)
PES#EE				0.1237 ***
				(4.8477)
EES#EE				−0.0301
				(−1.1215)
EE	−0.9198 ***	−0.8142 ***	−0.5942 ***	−0.4648 ***
	(−5.1169)	(−4.7938)	(−3.6338)	(−3.1039)
Year dummy	Yes	Yes	Yes	Yes
Industry dummy	Yes	Yes	Yes	Yes
Province dummy	Yes	Yes	Yes	Yes
Controls	Yes	Yes	Yes	Yes
N	6036.0000	6036.0000	6036.0000	6036.0000

Note: The symbols *** and ** denote significance at the 1% and 5% levels, respectively. Symbol “#” means variable interaction.

.

5. Conclusions and Policy Implications

This study systematically examines the impact of environmental subsidies (ES) on green innovation (GI) among heavy-polluting listed companies in China from 2012 to 2021, while investigating the moderating role of environmental enforcement (EE). The study yields several significant findings. First, ES generally inhibits GI; however, stringent EE actively promotes GI and can mitigate the adverse effects of ES. Second, different characteristics of ES demonstrate varying impacts on corporate GI. The negative influence primarily originates from ES beforehand, targeted ES, and pollution prevention ES. Strict EE can effectively compensate for these ES-related deficiencies, thereby fostering corporate GI. Furthermore, the study explores the allocation efficiency of ES and its interaction with EE through the lens of corporate ownership structure, environmental investment, and earnings management quality. The findings reveal that state-owned enterprises, companies lacking environmental investments, and those with higher earnings management levels demonstrate negative responses to ES. However, robust EE can transform these outcomes, indicating its potential to improve ES allocation issues and enhance overall GI levels among heavy-polluting enterprises.

Based on the findings, several policy implications emerge for China to promote the green transition of heavily polluting firms and achieve a sustainable economy. First, ES has proven insufficient in encouraging GI in these companies. Therefore, the government should strengthen the implementation of ES policies. Second, the design of ES must consider critical factors such as timing, form, and content, as these aspects significantly influence GI success. While many ES policies with more flexibility have shown greater efficiency in promoting GI. Thus, ES policies should be made more flexible and grant companies greater decision-making autonomy. Third, subsidy allocation directly impacts ES efficiency. Specifically, state-owned enterprises, companies lacking environmental protection foundations, and those obtaining subsidies through earnings management practices do not effectively utilize ES. Fourth, EE serves as a vital tool for addressing insufficient government investment in ES, as well as ES design and allocation challenges. The observed synergy between ES and EE measures strengthens the case for stricter environmental regulations in China. These suggestions offer valuable insights for other developing countries in their green transition phase. Subsidy-based policy tools require more rigorous design and must fully consider each country’s stage of green development. Most importantly, strict government enforcement remains crucial for ensuring the effectiveness of subsidy policies. In addition to providing recommendations for governments and policymakers, strategies targeted at internal managers within companies are equally important. Based on the results of our study, we offer the following specific advice to corporate managers: First, corporate managers should carefully evaluate the use of environmental subsidies and ensure that these funds are effectively allocated to promote green innovation projects rather than serving merely as short-term economic support. Second, companies should actively implement environmental protection measures to ensure that their production and business activities comply with current regulatory requirements. Third, managers should increase investment in environmental protection, viewing it as a crucial component of the company’s long-term development strategy, thereby enhancing the company’s green innovation capabilities. Fourth, managers need to review current financial management and reporting processes, aiming to improve transparency and accuracy. This will help in utilizing environmental subsidies more effectively to drive green innovation. Fifth, compared to non-state-owned enterprises, managers in state-owned enterprises should pay closer attention to how internal reforms and external collaborations can be used to enhance green innovation capabilities and overcome potential systemic barriers. This translation accurately conveys the original content, offering clear and actionable advice for corporate managers based on the study’s findings.

One limitation of our study is that we have not distinguished between innovation ES and non-innovation ES. The former refers to ES that is associated with innovation activities related to environmental protection, while the latter refers to ES that is related to non-innovation activities. By differentiating between these two types of ES, we could potentially identify different effects of ES on various externalities of GI. In particular, innovation ES may have a positive impact on both environmental protection externality and knowledge spillover externality, while non-innovation ES may primarily address environmental protection externality. Exploring this distinction will be an important topic for our future research. Meanwhile, ES and EE can influence each other. This study has investigated the moderating role of EE in the process ES affecting GI. Future research should further explore the results and mechanisms of how ES affects the effects of EE.