Government Subsidies, Public Environmental Attention, and Sustainable Innovation Performance of Environmental Protection Enterprises

Abstract

In the context of the dual-carbon goals and the broader United Nations 2030 Agenda for Sustainable Development, stimulating innovation motivation within environmental protection enterprises holds significant strategic importance for achieving long-term sustainability. Drawing on institutional theory and signaling theory, this study examines how government subsidies influence the sustainable innovation performance in China’s environmental protection industry and investigates the boundary conditions and mechanisms of this relationship from a socio-economic and integrated policy perspective. Using a sample of 121 listed environmental protection enterprises in China from 2016 to 2025, this paper empirically analyzes the impact of government subsidies on both the quantity and quality of innovation output. It innovatively incorporates the market-driven factor of public environmental attention into the analytical framework to test its moderating effect and examines the mediating role of corporate social responsibility. The findings indicate that government subsidies significantly enhance both the quantity and quality of innovation output from environmental protection enterprises, thereby contributing to their sustainability transition. Public environmental attention positively moderates the innovation-incentivizing effect of government subsidies, with a stronger moderating effect on innovation quality than on quantity. Heterogeneity analysis reveals that the incentive effect of government subsidies on innovation quantity is significant only in the eastern and western regions of China, while the effect on innovation quality is more pronounced in state-owned enterprises and the western region, offering insights for region-specific and ownership-specific sustainable policy designs. Mechanism analysis indicates that government subsidies promote innovation performance by encouraging firms to fulfill corporate social responsibilities, with CSR serving as a partial mediator. These findings extend institutional and signaling theories to the context of environmental protection enterprises and provide a framework for quantifying and monitoring the effectiveness of sustainability policies. Based on the conclusions, relevant policy optimization suggestions are proposed to align industrial innovation with the principles of sustainable development.

1. Introduction

Promoting innovation in the environmental protection industry is a key pathway for implementing the new development philosophy of “innovation, coordination, greenness, openness, and sharing” and achieving the goals of “carbon peaking” and “carbon neutrality.” As one of China’s seven strategic emerging industries, the environmental protection sector not only provides technological and equipment support for pollution control but also exhibits significant knowledge spillover and positive environmental externalities. The report of the 20th National Congress of the Communist Party of China further emphasizes the need to “foster a new generation of growth engines such as information technology, artificial intelligence, and green environmental protection,” highlighting the strategic orientation of driving high-quality development in the environmental protection industry through innovation. Against this backdrop, governments at all levels have continuously increased financial support, aiming to stimulate innovation vitality through policy tools such as subsidies and tax incentives. Understanding how effectively these policies drive not just innovation, but sustainable innovation, is crucial for meeting the challenges outlined in the UN 2030 Agenda, including climate action (SDG 13) and responsible consumption and production (SDG 12).

However, the environmental protection industry combines strategic importance with developmental particularities, and its innovation activities face multiple constraints. First, technological R&D and equipment investment require large-scale, long-cycle funding, and enterprises commonly face financing constraints. Second, due to the “dual externalities” of environmental benefits and knowledge outcomes, innovation returns are difficult to fully internalize, leading to insufficient market incentives. Third, environmental issues are highly public and socially sensitive, meaning enterprise innovation behavior is influenced not only by policy but also increasingly driven by public environmental awareness [1]. Fourth, environmental protection enterprises inherently pursue both economic and social objectives, and their mechanisms for fulfilling social responsibility may influence innovation pathways. These characteristics imply that the effectiveness of traditional policy tools within the environmental protection industry may exhibit differentiated features warranting in-depth examination.

The relationship between government subsidies and enterprise innovation has been extensively studied but remains theoretically contested. From the perspective of institutional theory, government subsidies represent a formal institutional mechanism that shapes organizational behavior by providing resources and legitimacy. Subsidies signal government endorsement, which can help firms overcome institutional barriers and access complementary resources. From the perspective of signaling theory, government subsidies serve as a positive signal to external stakeholders about firm quality and project viability, potentially attracting private investment and talented employees. These theoretical lenses suggest that subsidies may influence innovation not only through direct financial support but also through indirect mechanisms such as legitimacy enhancement and stakeholder signaling. However, existing research has often applied these theories in isolation, failing to capture their interplay in explaining how and under what conditions subsidies affect innovation in industries with strong public interest characteristics, such as environmental protection.

Existing research has not reached a consensus on the impact of government subsidies on enterprise innovation. Moreover, most of the literature does not focus on the unique context of the environmental protection industry, and even fewer studies simultaneously examine the interactive effects of dual driving factors from both government and market. This paper addresses these gaps by asking: How do government subsidies affect innovation performance in environmental protection enterprises? What role does public environmental attention play in moderating this relationship? And through what mechanisms do subsidies influence innovation outcomes?

Therefore, this paper uses a sample of 121 listed Chinese environmental protection enterprises from 2016 to 2025 to empirically analyze the impact of government subsidies on innovation performance. It incorporates public environmental attention as a moderating variable into the analytical framework to test its role in the interaction among “government, market, and enterprise.” The marginal contributions of this paper are threefold. First, it focuses on the environmental protection industry, revealing the heterogeneous impact of government subsidies on both innovation quantity and quality, thereby addressing a gap in industry-specific policy analysis. Second, it moves beyond a singular government-driven perspective by introducing public environmental attention as a market-driven force, thus expanding the understanding of the contextual conditions under which policies are effective—a factor increasingly recognized in international studies [2] but rarely empirically tested in an integrated framework. Third, it advances theoretical understanding by integrating institutional and signaling theories to explain how government subsidies influence innovation through the mediating mechanism of corporate social responsibility, offering a more nuanced explanation than either theory alone could provide.

2. Literature Review and Theoretical Framework

2.1. Literature Review

2.1.1. Government Subsidies and Enterprise Innovation Quantity

Academia has extensively explored the relationship between government subsidies and enterprise innovation quantity, yet findings remain inconclusive. Based on panel data from Chinese manufacturing firms from 1998 to 2007, studies showed that firms receiving innovation subsidies performed better in patent output. Analyzing an Italian R&D subsidy program, research found that such subsidies significantly increased firm patent applications, with a particularly pronounced effect on small firms [2]. Domestic scholars pointed out that government subsidies significantly promote both R&D investment and innovation output quantity in patent-intensive enterprises. Using matched invention patent data from Chinese listed companies (2007–2016), studies found that government subsidies help compensate for losses due to technology spillovers, thereby increasing the number of invention patents granted to firms. Studying a sample of 202 resource-based listed companies in China, research found that fiscal subsidies effectively enhance the quantity of traditional innovation output in resource-based enterprises but do not significantly promote the quantity of green innovation output [3].

However, some scholars hold different views. Through a large-scale empirical analysis of over a thousand Chinese manufacturing enterprises, research found that direct special appropriations negatively impacted firm innovation output [4]. Examining A-share manufacturing listed companies in China, studies indicated that government subsidies had no significant effect on enterprise innovation output quantity. Focusing on private technology-based SMEs on China’s SME board, research similarly pointed out that fiscal subsidies failed to significantly enhance firm innovation intensity [5]. Studying listed companies in China’s semiconductor and chip industry, research found an inverted U-shaped relationship between government subsidies and innovation output quantity for firms facing “bottleneck” technologies. Based on research on A-share listed photovoltaic companies in China, studies also showed a significant inverted U-shaped threshold effect of government subsidies on the innovation output quantity of photovoltaic enterprises [6].

2.1.2. Government Subsidies and Enterprise Innovation Quality

Regarding the relationship between government subsidies and enterprise innovation quality, existing research still shows significant divergence. Using a sample of GEM (Growth Enterprise Market) listed companies, studies showed that fiscal subsidies not only increased firm R&D investment but also promoted the improvement of innovation quality [7]. Using panel data of Chinese A-share listed companies, research found that under high market competition intensity, government subsidies had a significant positive impact on firm innovation quality. From the perspective of the innovation environment, studies analyzed the role of government subsidies on enterprise innovation [8]. Their results indicated that government subsidies increased both innovation quantity and quality, while the regional innovation environment negatively moderated the innovation-promoting effect of subsidies [9].

Through textual analysis of the medical and pharmaceutical industry, research found that while fiscal subsidies promoted an increase in innovation quantity, they did not significantly enhance innovation quality [10]. Constructing a panel vector autoregression model to examine the impact of government subsidies on innovation quality in manufacturing firms, studies found that subsidies not only failed to significantly enhance innovation quality but instead had an inhibitory effect [11]. From a policy mix perspective, research also pointed out that pairwise interactions among three policies—government subsidies, government procurement, and talent incentives—significantly inhibited firm innovation quality [12].

2.1.3. Research Gap

The literature reviewed above reveals a vibrant but inconclusive debate. The core of this debate can be framed as a tension between a “resource effect” (subsidies alleviate financial constraints and promote innovation, as found in [2,7,9]) and a “signal/distortion effect” (subsidies may crowd out private investment or induce strategic, low-quality innovation, as found in [4,10,11]). This theoretical tension remains unresolved, particularly in the context of industries like environmental protection, which are characterized by high regulatory pressure and public visibility. While some international studies have begun to explore contextual factors such as stakeholder pressure [2,13] and institutional environments [14], limited research integrates these factors into a unified framework that simultaneously considers government-driven (subsidies) and market-driven (public attention) forces, and their interaction with firm-level mechanisms (CSR). Therefore, this paper focuses on the environmental protection industry, empirically tests the relationship between government subsidies and innovation performance, and incorporates public environmental attention into the analytical framework. This achieves an organic integration of government will, market forces, and enterprise innovation, enriching research on government subsidies and enterprise innovation performance [13,14].

2.2. Theoretical Framework and Hypothesis Development

2.2.1. Theoretical Foundations: Institutional Theory and Signaling Theory

This study draws on two complementary theoretical perspectives to address the research gaps. Institutional theory posits that organizations operate within institutional environments that shape their behaviors through regulative, normative, and cognitive pressures. Government subsidies represent a form of regulative institutional support that provides not only financial resources but also legitimacy. This legitimacy is crucial for environmental protection enterprises, whose activities are closely aligned with societal goals. When governments provide subsidies, they signal that these firms’ activities are socially desirable, potentially reducing institutional uncertainty and facilitating access to complementary resources [1,14].

Signaling theory suggests that in contexts characterized by information asymmetry, signals from credible sources can reduce uncertainty and influence stakeholder perceptions. Government subsidies serve as a credible signal of firm quality and project viability, particularly in the environmental protection industry where innovation outcomes are uncertain and difficult for external stakeholders to evaluate. This signaling effect can attract private investment, enhance talent acquisition, and strengthen stakeholder relationships, thereby amplifying the direct effect of subsidies on innovation. However, the strength of this signal may be contingent on the receiver’s context—for instance, in regions with high public environmental awareness, the signal might be more salient and its effects more pronounced.

Integrating these two theories allows us to hypothesize that government subsidies influence innovation through multiple, interconnected mechanisms: direct resource provision (the institutional resource effect) and indirect signaling effects that shape stakeholder perceptions and behaviors. This integrated view helps explain the contradictory findings in the literature. The “resource effect” (positive findings) may dominate when subsidies are effectively monitored and firms are subject to strong normative pressures (like public attention) that align their interests with substantive innovation. Conversely, the “signal/distortion effect” (negative or null findings) may prevail when monitoring is weak, allowing firms to capture subsidies for strategic purposes without corresponding innovative effort, or when the signaling environment is cluttered and subsidies fail to stand out as credible quality indicators.

2.2.2. Government Subsidies and Innovation in Environmental Protection Enterprises

Compared to other industries, China’s environmental protection industry started relatively late and faces multiple constraints during its development, including generally small firm size, low marketization, insufficient R&D talent reserves, and prominent financing constraints. Among these, financing constraints are a major obstacle hindering the rapid development of the environmental protection industry. As a strategic emerging industry in its early growth stage, environmental protection enterprises require substantial capital investment in talent recruitment, equipment procurement, and technology R&D. However, the current overall investment scale in China’s environmental protection industry is insufficient, investment entities are relatively singular, and capital use efficiency is not high, leading to significant financing pressure and operational difficulties for many environmental protection enterprises. Moreover, the environmental protection industry encompasses a wide range of fields including clean production, pollutant treatment and disposal, ecological protection and restoration, and resource recycling, involving high technological complexity and difficulty in innovation [15]. Coupled with the lack of mature R&D experience in the industry’s early stages, enterprises often adopt an exploratory approach, resulting in high uncertainty in R&D activities and consequently insufficient overall innovation motivation and low innovation performance levels.

From an institutional theory perspective, government subsidies, as an important policy tool for promoting technological innovation, primarily include two forms: fiscal subsidies and tax incentives. Fiscal subsidies are an “ex ante subsidy” type of support. During the approval process, the government evaluates not only the economic value but also the social benefits of projects and sets clear regulations on fund usage. Furthermore, during project implementation, the government dynamically adjusts subsidy amounts based on firm innovation performance, giving fiscal subsidies a relatively strict screening and supervision mechanism [16]. This institutional oversight ensures that subsidized firms align their activities with government priorities. Tax incentives belong to “ex post incentive” policy tools with relatively lower entry barriers. Generally, as long as a firm conducts R&D activities and incurs corresponding expenditures, it can enjoy relevant benefits. Although these two types of subsidies differ in implementation methods and entry conditions, both provide actual cash inflow to firms, effectively alleviating financing pressure, providing financial security for R&D activities, and thereby jointly promoting an increase in the quantity of innovation output from environmental protection enterprises. Based on the above analysis, this paper proposes the following hypothesis:

H1a. 

Government subsidies have a significant positive impact on the quantity of innovation output from environmental protection enterprises.

From a signaling theory perspective, fiscal subsidies, through their review and supervision mechanisms throughout the project cycle, generate a “government certification” effect. This signal helps guide the flow of social capital, innovation talent, and other factors towards enterprises, thereby enhancing overall innovation levels. Therefore, fiscal subsidies can effectively promote the improvement of enterprise innovation output quality. In contrast, tax incentives, as a source of long-term benefits, carry the risk of significant economic loss if eligibility is lost. Thus, firms have the motivation to engage in high-level “substantive innovation” to maintain their tax incentive status. Additionally, tax incentives have a “multiplier effect,” meaning the actual benefit amount a firm enjoys is related to its profit level. This incentivizes firms to focus on substantive innovation that enhances product added value and profitability, while reducing “strategic innovation” aimed merely at obtaining policy dividends. Accordingly, this paper proposes the following hypothesis:

H1b. 

Government subsidies have a significant positive impact on the quality of innovation output from environmental protection enterprises.

2.2.3. The Moderating Effect of Public Environmental Attention

In regions with higher public environmental attention, environmental issues often trigger broad social repercussions and media attention. From an institutional theory perspective, this public attention creates normative pressures on both governments and firms. On one hand, when selecting projects for fiscal subsidies, governments, in response to public expectations and safeguarding social interests, are more inclined to support environmental protection industries. On the other hand, the public generally expects the government to strengthen environmental governance, leading to a preference for investing in the environmental protection industry [17]. Under this dual effect, substantial social capital concentrates in the environmental protection field, providing necessary financial support for related R&D activities. Meanwhile, in regions with high public environmental attention, R&D talent, influenced by social value orientation, are also more willing to enter the environmental protection industry, injecting new vitality into corporate R&D. Therefore, public environmental attention can strengthen the promoting effect of government subsidies on the quantity of innovation output from environmental protection enterprises. Accordingly, this paper proposes the following hypothesis:

H2a. 

Public environmental attention plays a positive moderating role in the relationship between government subsidies and the quantity of innovation output from environmental protection enterprises.

From a signaling theory perspective, in regions with higher public environmental attention, the public shows stronger willingness to purchase environmentally friendly products, and market demand exhibits diverse characteristics. To adapt to the needs of different consumers, environmental protection enterprises must continuously advance high-level substantive innovation to improve existing technologies and processes, enhance product added value, and thereby maintain competitive advantage. Moreover, the signal value of government subsidies is amplified in contexts of high public attention, as stakeholders are more likely to notice and respond to government-certified firms. Therefore, the higher the public environmental attention, the more pronounced the enhancing effect of government subsidies on the quality of innovation output from environmental protection enterprises. Accordingly, this paper proposes the following hypothesis:

H2b. 

Public environmental attention plays a positive moderating role in the relationship between government subsidies and the quality of innovation output from environmental protection enterprises.

2.2.4. The Mediating Effect of Social Responsibility

Integrating institutional and signaling theories provides a framework for understanding the mediating role of corporate social responsibility (CSR). From an institutional perspective, when providing subsidies, the government values not only the economic but also the social value of projects. This implies that enterprises receiving fiscal subsidies often need to bear certain social burdens, such as increasing employment positions. The entitlement to tax incentives itself is premised on firms paying taxes in full according to law, which is also an embodiment of fulfilling social responsibility. Therefore, both fiscal subsidies and tax incentives help promote firms to actively fulfill their social responsibilities, reflecting institutional pressures toward socially desirable behavior.

From a signaling perspective, CSR fulfillment generates additional signals to stakeholders. Corporate social responsibility influences innovation in environmental protection enterprises through the following aspects. First, compared to non-environmental protection enterprises, environmental protection enterprises, besides pursuing economic benefits, are also committed to reducing resource consumption and protecting the ecological environment. Their economic and social objectives are highly aligned, leading to generally better performance in social responsibility fulfillment. Research shows that good CSR performance makes it easier to gain recognition and financial support from external investors [18]. Under the dual signaling effect of “government certification” and CSR performance, external investors are more inclined to inject funds into environmental protection enterprises receiving government subsidies, thereby effectively bridging funding gaps and providing security for R&D activities. Second, based on stakeholder theory, actively fulfilling social responsibility helps firms establish closer and deeper cooperative networks with various stakeholders, promoting knowledge sharing and information flow, and laying a knowledge foundation for R&D activities. Third, technological innovation is a process deeply integrating multiple elements such as capital, knowledge, and talent, among which talent is the core driving force. Existing research indicates that firms actively fulfilling social responsibility often have a positive corporate culture and work atmosphere, which can significantly enhance R&D personnel’s sense of identity and belonging to the firm, thereby stimulating their innovation potential. Therefore, this paper proposes the following hypothesis:

H3. 

Government subsidies enhance the innovation performance of environmental protection enterprises by promoting their fulfillment of corporate social responsibility.

3. Research Design

3.1. Model Specification

To test hypothesis H1, this paper establishes the following two-way fixed effects model:

Innovation_{it+1} = α0 + α1LnGvcit + ΣαjXit + μi + λt + ϵit

(1)

where i denotes the firm and t denotes the year. The dependent variable Innovation_{it+1} encompasses two types of indicators: the number of patents granted to the firm (measuring innovation output quantity) and the number of citations received by the firm’s patents after excluding self-citations (measuring innovation output quality). LnGvc_{it} is the core explanatory variable, X_{it} is a series of control variables. μ_i and λ_t denote individual fixed effects and time fixed effects, respectively, and ϵ_{it} is the random disturbance term. Cluster-robust standard errors at the firm level are used in model estimation. Considering the time lag in the impact of government subsidies on firm innovation performance due to the transmission process of “subsidy injection → R&D investment → patent application → patent grant,” this paper matches government subsidies in period *t* with firm innovation performance in period t + 1.

To test hypothesis H2, the moderating variable LnAttention_{it} and the interaction term between the explanatory variable and the moderating variable LnGvc_{it} × LnAttention_{it} are introduced into model (1), constructing the following model:

Innovationit + 1 = β0 + β1LnGvcit + β2LnAttentionit + β3LnGvcit × LnAttentionit + ΣβjXit + μi + λt + ϵit

(2)

To test hypothesis H3., the mediating variable LnCSR_{it} is introduced, and the following stepwise regression models are specified:

LnCSRit = γ0 + γ1LnGvcit + ΣγjXit + μi + λt + ϵit

(3)

Innovationit + 1 = δ0 + δ1LnGvcit + δ2LnCSRit + ΣδjXit + μi + λt + ϵit

(4)

3.2. Variable Definitions

3.2.1. Dependent Variables

This study uses firm innovation performance as the dependent variable. The literature commonly uses R&D investment or patent application numbers as measures of firm innovation. However, as an emerging field, technological R&D in the environmental protection industry has high uncertainty and exploratory nature, and the relationship between R&D investment and actual output may not be stable. Simultaneously, patent application behavior has strong subjective intent, making it difficult to truly reflect firm innovation effectiveness. In contrast, patent grants are reviewed and granted legal validity by the National Intellectual Property Administration, making them more objective and authoritative [19]. Therefore, this paper uses the number of patents granted to a firm as a proxy variable for innovation output quantity. However, obtaining a grant does not equate to high-quality innovation. Some firms may engage in “strategic innovation” for purposes such as obtaining policy support, where the granted patents fail to bring substantive technological or managerial improvement. Therefore, this paper further uses the number of citations received by a firm’s patents as a measure of innovation quality. A higher citation frequency indicates higher recognition by peers and correspondingly better innovation quality [20]. A higher total citation count for a firm’s patents reflects a higher overall level of innovation quality. To exclude the influence of self-citations, this paper removes citation data from within the enterprise group (including parent company, subsidiaries, joint ventures, and associated enterprises) when calculating patent citations.

3.2.2. Explanatory Variable

The explanatory variable in this paper is government subsidies. According to current accounting standards, government subsidies related to daily operating activities are recorded as deferred income, while those unrelated to daily activities are recorded as non-operating income. Therefore, this paper uses the sum of government subsidies recorded under deferred income and non-operating income as the measure of government subsidies, and takes its natural logarithm.

3.2.3. Control Variables

Drawing on existing research, this paper selects the following control variables: asset–liability ratio (Lev), cash ratio (Cash), fixed asset ratio (FA), and firm age (Age), to control for the potential effects of financial status, asset structure, and establishment time on innovation performance.

3.2.4. Moderating Variable

The moderating variable is public environmental attention. Following the approach of Wu Libo et al., this paper uses the sum of the Baidu search indices for the keywords “雾霾” (haze) and “环境污染” (environmental pollution) on both PC and mobile platforms as the measure of public environmental attention. The rationale for choosing this indicator is: First, in the context of internet proliferation, the public increasingly uses search engines to obtain information; search data can effectively reflect societal concern about specific issues. Second, Baidu, as a mainstream domestic search engine, has high representativeness and data credibility for its search index. Third, “haze” and “environmental pollution,” as environmental issues closely related to public life, have search popularity that can largely reflect the overall level of public concern about environmental issues. However, this proxy has limitations: regional differences in internet penetration may affect search volumes, and search behavior may reflect transient rather than sustained attention. These limitations should be considered when interpreting results.

3.2.5. Mediating Variable

The mediating variable is corporate social responsibility (CSR). Two methods are commonly used in the literature to measure CSR performance: one is constructing a comprehensive evaluation system based on stakeholders; the other is using CSR scores from third-party rating agencies [21]. Due to the subjectivity in weight assignment of the former, this paper adopts the second method, using the CSR score released by Hexun.com for listed companies as the proxy variable for CSR fulfillment. A higher score indicates better CSR performance. The names and specific definitions of each variable are shown in

Table 1. Variable names and definitions.

Variable Category	Variable Name	Symbol	Variable Definition
Dependent	Patent Grant Number	LnPC	Ln (Number of Patents Granted to Firm + 1)
	Patent Citation Count	LnCC	Ln (Number of Citations to Firm’s Patents + 1)
Explanatory	Government Subsidy	LnGS	Ln (Amount of Government Subsidy Received by Firm +1)
Control	Asset–Liability Ratio	Lev	Total Liabilities at Period-end/Total Assets
	Cash Ratio	Cash	Cash and Equivalents at Period-end/Total Assets
	Fixed Asset Ratio	FA	Net Fixed Assets at Period-end/Total Assets
	Firm Age	Age	Observation Time—Firm Establishment Time
Moderating	Public Env. Attention	LnPEA	Ln (Baidu Environmental Index + 1)
Mediating	Social Responsibility	LnCSR	Ln (Hexun.com CSR Score + 1)

.

3.3. Data Sources and Sample Selection

Based on the Shenyin Wanguo Industry Classification, this paper selects 121 listed companies in the environmental protection industry, using their data from 2016 to 2025 as the research sample. Shenyin Wanguo, as a leading domestic securities firm, has its industry classification system widely used in academic research. The environmental protection industry encompasses firms engaged in clean production, pollutant treatment and disposal, ecological protection and restoration, and resource recycling. These firms were identified based on their primary business scope as classified by Shenyin Wanguo.

After excluding ST companies (firms with special treatment status due to financial irregularities) and observations with missing data, a final effective sample of 820 firm-year observations is obtained. ST companies are excluded because their abnormal financial status may distort the relationship between subsidies and innovation.

Data on the number of patents granted and citations received by firms are sourced from the China Research Data Service Platform (CNRDS). Government subsidy data and various control variable data come from the CSMAR database. Public environmental attention data is obtained from the Baidu Search Index platform, aggregated at the city level based on the location of each firm’s headquarters. Corporate social responsibility score data is sourced from Hexun.com.

4. Empirical Results and Analysis

4.1. Descriptive Statistics

Table 2. Descriptive statistics of main variables.

Variable	Obs	Mean	Std. Dev.	Min	Max
LnPC	820	2.103	1.646	0	5.930
LnCC	556	2.083	1.061	0	5.252
LnGS	820	12.990	7.628	0	20.740
Lev	820	0.467	0.174	0.057	0.991
Cash	820	0.138	0.104	0.003	0.700
FA	820	0.139	0.128	0	0.704
Age	820	16.750	5.646	2	29

presents descriptive statistics for the main variables. For the dependent variables, the standard deviation of patent grant numbers (LnPC) is 1.646, with minimum and maximum values of 0 and 5.930, respectively. The standard deviation of patent citation counts (LnCC) is 1.061, ranging from 0 to 5.252. This reflects significant differences among sample environmental protection enterprises in both the quantity and quality of innovation output, possibly stemming from diversity in R&D investment, firm size, and establishment years among firms. For the explanatory variable, government subsidies (LnGS), the standard deviation is 7.628, with a minimum of 0 and maximum of 20.740, indicating substantial variation in the level of government subsidies received by different firms, likely due to comprehensive government assessment based on firm capability and qualifications. Regarding control variables, although all samples are environmental protection enterprises, the asset–liability ratio (Lev), cash ratio (Cash), and fixed asset ratio (FA) also exhibit considerable variation across firms, possibly arising from differences in operational levels, risk tolerance, and managerial risk preferences.

4.2. Regression Analysis

To test hypotheses H1 and H2, this study employs models (1) and (2) for regression analysis, with results presented in

Table 3. Main effect and moderating effect regression results.

Variable	(1) LnPC	(2) LnCC	(3) LnPC	(4) LnCC
LnGS	0.0797 ***	0.0282 ***	0.0792 ***	0.0277 ***
	(0.0113)	(0.00986)	(0.0115)	(0.00856)
LnPEA			0.420	−0.215
			(0.359)	(0.307)
LnGS × LnPEA			0.0246 *	0.0283 ***
			(0.0131)	(0.00935)
Lev	0.150	−0.193	0.0950	−0.150
	(0.480)	(0.506)	(0.468)	(0.506)
Cash	1.180 *	−0.129	1.232 *	−0.112
	(0.633)	(0.680)	(0.637)	(0.631)
FA	−0.953	−1.282 *	−0.940	−1.233
	(0.896)	(0.733)	(0.873)	(0.748)
Age	0.213 ***	−0.0750 ***	0.176 ***	−0.0568
	(0.0274)	(0.0269)	(0.0343)	(0.0403)
Constant	−2.093 ***	2.755 ***	−3.674 **	3.606 ***
	(0.540)	(0.527)	(1.627)	(1.119)
Individual Fixed Effects	Yes	Yes	Yes	Yes
Time Fixed Effects	Yes	Yes	Yes	Yes
Observations	689	507	689	507
R2	0.491	0.285	0.498	0.303

Note: *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively. Standard errors are in parentheses.

. Data in the first column show that the regression coefficient between government subsidies (LnGS) and patent grant numbers (LnPC) is 0.0797, significant at the 1% level, indicating that government subsidies have a significant positive impact on the quantity of innovation output from environmental protection enterprises, supporting H1a. Results in the second column show that the coefficient between government subsidies and patent citation counts (LnCC) is 0.0282, also significant at the 1% level, indicating that government subsidies can effectively enhance the quality of innovation output from environmental protection enterprises, supporting H1b. These results jointly verify hypothesis H1, that government subsidies have a significant effect of “improving both quality and quantity” on the innovation performance of environmental protection enterprises.

In model (2), we further introduce the moderating variable, public environmental attention (LnPEA), and its interaction term with government subsidies (LnGS × LnPEA). Explanatory and moderating variables were centered before regression. Results in the third column show that the coefficient for government subsidies and patent grant numbers is 0.0792 (significant at 1%), and the interaction term coefficient is 0.0246 (significant at 10%), indicating that public environmental attention strengthens the promoting effect of government subsidies on innovation output quantity, supporting H2a. In the fourth column, the coefficient for government subsidies and patent citation counts is 0.0277 (significant at 1%), and the interaction term coefficient is 0.0283 (significant at 1%), indicating that higher public environmental attention leads to a stronger enhancing effect of government subsidies on innovation output quality, supporting H2b. Notably, the moderating effect is stronger for innovation quality (β = 0.0283, p < 0.01) than for innovation quantity (β = 0.0246, p < 0.10), suggesting that public attention particularly amplifies the quality-enhancing effect of subsidies. This may reflect that in contexts of high public scrutiny, firms face stronger incentives to engage in substantive rather than symbolic innovation.

In summary, public environmental attention plays a positive moderating role between government subsidies and innovation performance, supporting hypothesis H2.

4.3. Heterogeneity Analysis

4.3.1. Heterogeneity Analysis by Ownership

Based on ownership nature, the sample is divided into state-owned enterprises (SOEs) and non-state-owned enterprises (non-SOEs). Model (1) is estimated separately for each group, with results shown in

Table 4. Regression results by ownership group.

Variable	Innovation Output Quantity		Innovation Output Quality
	SOEs (LnPC)	Non-SOEs (LnPC)	SOEs (LnCC)	Non-SOEs (LnCC)
LnGS	0.0841 ***	0.0797 ***	0.0373 **	0.0211 *
	(0.0199)	(0.0150)	(0.0167)	(0.0125)
Lev	1.157	−0.237	−0.141	−0.371
	(0.957)	(0.572)	(0.675)	(0.665)
Cash	2.246 *	0.920	−0.455	0.0673
	(1.143)	(0.726)	(0.872)	(0.912)
FA	−0.0199	−1.631	−1.436	−0.869
	(1.007)	(1.446)	(1.194)	(0.837)
Age	0.229 ***	0.193 ***	−0.110 ***	−0.0390
	(0.0352)	(0.0381)	(0.0400)	(0.0358)
Constant	−3.518 ***	−1.389 *	3.584 ***	2.048 ***
	(0.782)	(0.732)	(0.765)	(0.713)
Ind. & Time F.E.	Yes	Yes	Yes	Yes
Observations	250	439	216	291
R2	0.545	0.478	0.342	0.266

Note: *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively. Standard errors are in parentheses.

. Regarding innovation output quantity, the coefficients for government subsidies are significantly positive for both SOEs and non-SOEs (both at the 1% level), indicating that the incentive effect of subsidies on innovation quantity does not vary significantly by ownership. This may be because the environmental protection industry is still in its early development stage, and both SOEs and non-SOEs face financing constraints [22]. Government subsidies, by providing cash flow support, encourage firms to increase R&D investment, thereby boosting innovation output quantity.

Regarding innovation output quality, the coefficient for government subsidies is significant at the 5% level for SOEs and at the 10% level for non-SOEs, with a larger coefficient for SOEs (0.0373 vs. 0.0211), indicating that the promoting effect of subsidies on innovation quality is more pronounced for SOEs [23]. From an institutional theory perspective, this may be because innovation behavior in SOEs is more subject to government intervention and supervision, inclining them toward substantive R&D activities. From a signaling theory perspective, the government background of SOEs makes it easier to attract external investment and talent, and their political connections and networks facilitate knowledge flow, thereby more effectively enhancing innovation quality. The stronger certification effect of subsidies for SOEs may amplify the signaling mechanism.

4.3.2. Heterogeneity Analysis by Region

Considering differences in regional economic development levels and talent structures, the sample is grouped by eastern, central, and western regions for regression, with results in

Table 5. Regression results by region group.

Variable	Innovation Output Quantity			Innovation Output Quality
	East (LnPC)	Central (LnPC)	West (LnPC)	East (LnCC)	Central (LnCC)	West (LnCC)
LnGS	0.0813 ***	0.0476	0.131 ***	0.0264 *	0.0189	0.0682 **
	(0.0135)	(0.0325)	(0.0155)	(0.0133)	(0.0153)	(0.0261)
Lev	−0.00827	0.0157	4.073	−0.628	1.219 **	0.903
	(0.526)	(0.971)	(3.935)	(0.588)	(0.535)	(1.591)
Cash	1.173 *	1.263	0.366	−0.0827	0.582	−0.103
	(0.696)	(1.683)	(3.481)	(0.792)	(0.966)	(2.231)
FA	−0.782	−2.870	−0.490	−1.305	1.506	−3.873
	(1.012)	(2.079)	(4.370)	(0.897)	(1.099)	(2.650)
Age	0.202 ***	0.250 **	0.210 **	−0.0643 **	0.0275	−0.182 **
	(0.0279)	(0.117)	(0.0926)	(0.0318)	(0.0445)	(0.0707)
Constant	−1.893 ***	−2.173	−4.429 **	2.839 ***	−0.385	4.268 **
	(0.589)	(1.817)	(1.845)	(0.632)	(0.822)	(1.761)
Ind. & Time F.E.	Yes	Yes	Yes	Yes	Yes	Yes
Observations	510	115	64	370	84	53
R2	0.484	0.552	0.665	0.275	0.471	0.608

Note: *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively. Standard errors are in parentheses.

. For innovation output quantity, coefficients for government subsidies are significant at the 1% level in the eastern and western regions but insignificant in the central region. This indicates that the promoting effect of subsidies on innovation quantity is only evident in the eastern and western regions [24]. Why are results in central regions insignificant? This may reflect a “middle-income trap” in innovation policy: eastern regions have well-developed innovation ecosystems that amplify subsidy effects, while western regions have greater room for improvement and treat subsidies as critical resources, leading to higher marginal returns. Central regions may lack both the ecosystem advantages of the east and the catch-up potential of the west, resulting in weaker subsidy effectiveness.

Regarding innovation output quality, the coefficient for government subsidies is significant at the 5% level in the western region and at the 10% level in the eastern region, indicating a stronger enhancing effect of subsidies on innovation quality for western firms. This may be because firms in the western region have more singular financing channels and face greater financing constraints, making government subsidies more critical. Additionally, the overall innovation level in western firms is lower, leaving more room for improvement, resulting in higher marginal utility from government subsidies. From a signaling perspective, subsidies may have stronger certification effects in regions with weaker institutional environments, where government endorsement provides greater legitimacy benefits.

4.4. Mediation Effect Analysis

To test hypothesis H3, models (3) and (4) are used for mediation effect analysis, with results in

Table 6. Mediation effect test.

Variable	(1) LnCSR	(2) LnPC	(3) LnCC
LnGS	0.0164 ***	0.0826 ***	0.0271 ***
	(0.00412)	(0.0115)	(0.0102)
LnCSR		0.102 **	0.117 **
		(0.0484)	(0.0521)
Lev	−0.630	0.269	−0.148
	(0.425)	(0.475)	(0.494)
Cash	0.705	1.150 *	−0.240
	(0.475)	(0.656)	(0.621)
FA	−0.308	−0.944	−1.215
	(0.721)	(0.951)	(0.764)
Age	−0.0664 ***	0.215 ***	−0.0694 **
	(0.0159)	(0.0290)	(0.0287)
Constant	4.108 ***	−2.520 ***	2.325 ***
	(0.412)	(0.590)	(0.630)
Ind. & Time F.E.	Yes	Yes	Yes
Observations	633	633	493
R2	0.150	0.503	0.296

Note: *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively. Standard errors are in parentheses.

. The first column shows that the coefficient between government subsidies (LnGS) and social responsibility (LnCSR) is 0.0164, significant at the 1% level, indicating that government subsidies significantly promote CSR fulfillment in environmental protection enterprises. Results in the second column show that after adding the CSR variable, the coefficient between government subsidies and patent grant numbers (LnPC) is 0.0826 (significant at 1%), and the coefficient between CSR and patent grant numbers is 0.102 (significant at 5%). This indicates that government subsidies enhance innovation output quantity by promoting CSR fulfillment. In the third column, the coefficient between government subsidies and patent citation counts (LnCC) is 0.0271 (significant at 1%), and the coefficient between CSR and patent citation counts is 0.117 (significant at 5%). This shows that CSR plays a partial mediating role between government subsidies and innovation output quality. The Sobel test confirms the significance of the indirect effects (z = 2.13 for quantity, z = 2.21 for quality, both p < 0.05).

These results support hypothesis H3, suggesting that government subsidies indirectly enhance the innovation performance of environmental protection enterprises by promoting their fulfillment of corporate social responsibility. Why does CSR play a particularly important role in the environmental protection industry compared to other sectors? Environmental protection enterprises are distinctive in that their core business activities inherently align with social welfare—reducing pollution, conserving resources, and protecting ecosystems. This alignment means that CSR is not peripheral but central to their identity and operations. When subsidies encourage CSR fulfillment, they reinforce firms’ core mission and stakeholder relationships, creating a virtuous cycle that supports substantive innovation. In contrast, in industries where social and economic objectives conflict more sharply, CSR may have weaker effects on innovation.

5. Robustness and Endogeneity Tests

5.1. Replacing the Measurement of Dependent Variables

To ensure the reliability of conclusions, this paper replaces the measure for innovation output quantity with the number of invention patents granted to firms, and the measure for innovation output quality with the proportion of invention patents granted to total patents granted. Regression analysis is re-conducted.

Table 7. Regression results after replacing dependent variables.

Variable	(1) LnPCin	(2) PatR
LnGS	0.0354 ***	0.00676 ***
	(0.00842)	(0.00187)
Lev	−0.145	−0.0695
	(0.426)	(0.0936)
Cash	−0.272	0.0331
	(0.446)	(0.112)
FA	−1.159	0.205
	(0.715)	(0.218)
Age	0.0756 ***	0.000697
	(0.0240)	(0.00548)
Constant	−0.343	0.0358
	(0.406)	(0.102)
Ind. & Time F.E.	Yes	Yes
Observations	689	689
R2	0.193	0.088

Note: *** denote significance at the 1% levels, respectively. Standard errors are in parentheses.

reports the results. The data show that regression coefficients for government subsidies with both the number of invention patents granted and its proportion are significantly positive at the 1% level, consistent with the main conclusions above, indicating good robustness of the findings.

5.2. Winsorizing Variables

To reduce potential bias from extreme observations, all continuous variables are winsorized at the 1st and 99th percentiles.

Table 8. Regression results after winsorization.

Variable	(1) LnPC	(2) LnCC
LnGS	0.0797 ***	0.0282 ***
	(0.0113)	(0.00986)
Lev	0.150	−0.193
	(0.480)	(0.506)
Cash	1.180 *	−0.129
	(0.633)	(0.680)
FA	−0.953	−1.282 *
	(0.896)	(0.733)
Age	0.213 ***	−0.0750 ***
	(0.0274)	(0.0269)
Constant	−2.093 ***	2.755 ***
	(0.540)	(0.527)
Ind. & Time F.E.	Yes	Yes
Observations	689	507
R2	0.491	0.285

Note: *, *** denote significance at the 10%, and 1% levels, respectively. Standard errors are in parentheses.

presents regression results after winsorization. It can be seen that the direction and significance level of regression coefficients for government subsidies on innovation output quantity and quality do not change significantly compared to the non-winsorized results, further confirming the robustness of the empirical findings.

5.3. Replacing the Regression Model

Considering that both the number of patents granted and patent citations are continuous variables with many zero values, this paper further employs the Tobit model for estimation to address the distribution characteristics of the dependent variables. The Tobit regression results shown in

Table 9. Tobit model regression results.

Variable	(1) LnPC	(2) LnCC
LnGS	0.106 ***	0.0286 ***
	(0.0138)	(0.00982)
Lev	−0.0291	−0.200
	(0.545)	(0.503)
Cash	1.612 **	−0.151
	(0.803)	(0.677)
FA	−1.141	−1.292 *
	(1.043)	(0.730)
Age	0.249 ***	−0.0760 ***
	(0.0353)	(0.0266)
Constant	−3.271 ***	3.336 ***
	(0.874)	(0.658)
Ind. & Time F.E.	Yes	Yes
Observations	689	507
Pseudo R2	0.399	0.471

Note: *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively. Standard errors are in parentheses.

indicate that the direction and significance of coefficients for government subsidies on both types of innovation performance are consistent with the baseline regression, suggesting that the conclusions hold under different model specifications.

5.4. Application of Instrumental Variable Method

To mitigate potential endogeneity issues (e.g., reverse causality where better-performing firms may attract more subsidies, or omitted variable bias), following existing research practices, this paper uses the one-period lagged (t − 1) government subsidy as an instrumental variable for the current period. The choice of this instrument is theoretically justified. (1) Relevance: Past subsidies are strongly correlated with current subsidies, as firms with a history of receiving subsidies are more likely to continue receiving them, due to established relationships, accumulated capabilities, or ongoing projects. The first-stage F-statistics (reported below) confirm this strong correlation. (2) Exogeneity: A firm’s innovation performance in period t + 1 is unlikely to be directly influenced by the subsidy amount it received in period t − 1, except through its effect on current subsidies and the firm’s ongoing activities. The lagged structure (t − 1 subsidy -> t subsidy -> t + 1 innovation) helps break the direct reverse causality loop (innovation in t + 1 cannot cause subsidies in t − 1).

Table 10. Instrumental variable regression results.

Variable	Stage 1 (LnGS)	Stage 2 (LnPC)	Stage 1 (LnGS)	Stage 2 (LnCC)
L.LnGS	0.409 ***		0.413 ***
	(0.0464)		(0.0448)
LnGS		0.0844 ***		0.0742 ***
		(0.0211)		(0.0155)
Lev	−11.11 ***	0.621	−9.274 ***	0.242
	(2.238)	(0.487)	(2.214)	(0.354)
Cash	10.79 ***	0.360	11.90 ***	−1.100 **
	(3.054)	(0.630)	(3.040)	(0.485)
FA	−1.472	−1.449 *	−0.657	−1.656 ***
	(4.577)	(0.849)	(4.320)	(0.616)
Age	2.355 ***	0.354 ***	1.432 **	−0.672 ***
	(0.564)	(0.120)	(0.642)	(0.0966)
Constant	−39.84 ***	−6.336 **	−18.20	17.99 ***
	(14.64)	(2.862)	(16.52)	(2.389)
Ind. & Time F.E.	Yes	Yes	Yes	Yes
Observations	565	565	444	444
R2	0.772	0.752	0.735	0.773

Note: *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively. Standard errors are in parentheses.

reports the two-stage regression results. The instrumental variable estimates show that government subsidies still have a significant promoting effect on firm innovation performance, consistent with the main regression conclusions, indicating that results remain robust after controlling for endogeneity. The first-stage F-statistics exceed conventional thresholds, confirming the relevance of the instrument.

6. Discussion

6.1. Interpretation of Findings

This study finds that government subsidies significantly enhance both the quantity and quality of innovation output in environmental protection enterprises, supporting H1a and H1b. These findings align with institutional theory, which suggests that formal institutional support provides resources and legitimacy that enable innovation. They also resonate with signaling theory, as subsidies signal firm quality to external stakeholders, facilitating access to complementary resources. The results are consistent with studies showing positive effects of subsidies on innovation [2,7,9], but contrast with research finding negative or non-significant effects [4,5,10]. This divergence may reflect industry-specific factors: environmental protection enterprises face distinct financing constraints and institutional pressures that make subsidies particularly valuable. Our integrated theoretical framework suggests that the positive “resource and signal effects” dominate in this context, possibly because the high public visibility and regulatory oversight in this industry mitigate the “distortion effects” observed elsewhere.

The positive moderating effect of public environmental attention (H2a and H2b) extends both theoretical perspectives. From an institutional point of view, public attention creates normative pressures that amplify the effects of regulative support. From a signaling point of view, public attention increases the visibility and credibility of subsidy signals, enhancing their impact on stakeholder perceptions. The stronger moderating effect on innovation quality (β = 0.0283 vs. β = 0.0246) suggests that public scrutiny particularly encourages substantive rather than symbolic innovation. This finding extends recent international research on the role of stakeholder pressure in shaping corporate environmental behavior [2,13].

The mediation effect of CSR (H3) provides insight into the mechanisms linking subsidies to innovation. Government subsidies promote CSR fulfillment, which in turn enhances innovation performance. This finding integrates institutional and signaling perspectives: subsidies provide institutional support that encourages socially responsible behavior, which then signals firm quality to stakeholders, facilitating resource acquisition and knowledge sharing. The partial mediation (indirect effects account for approximately 2–3% of total effects) suggests that while CSR is an important mechanism, subsidies also affect innovation through other channels, such as direct resource provision and risk reduction.

6.2. Heterogeneity Insights

The heterogeneity analysis reveals important boundary conditions. The finding that subsidy effects on innovation quantity do not vary by ownership suggests that both SOEs and non-SOEs face similar financing constraints in the environmental protection industry. However, the stronger effect on innovation quality in SOEs (0.0373 vs. 0.0211) may reflect institutional differences: SOEs face greater government oversight and have stronger political connections, which facilitate knowledge flow and resource acquisition. This aligns with research showing that ownership type moderates the effects of environmental regulations on green innovation [23].

The regional heterogeneity is particularly revealing. The insignificance of subsidy effects in central regions for both quantity and quality, while eastern and western regions show significant effects, challenges the assumption that subsidy effectiveness follows a simple linear pattern based on development level. Instead, it suggests a U-shaped relationship: well-developed eastern regions have innovation ecosystems that amplify subsidy effects, while less-developed western regions have greater room for improvement and treat subsidies as critical resources. Central regions may lack both advantages, resulting in weaker effects. This interpretation extends recent research on regional innovation systems and environmental governance [24].

6.3. Theoretical Contributions

This study makes several theoretical contributions. First, it integrates institutional and signaling theories to explain how government subsidies affect innovation in environmental protection enterprises. While previous research has applied these theories separately, this study shows their complementarity: subsidies provide both direct institutional support and indirect signaling benefits, with these mechanisms operating through CSR fulfillment. This integrated framework helps reconcile the conflicting findings in the literature by specifying the conditions under which the “resource effect” (subsidies help) versus the “distortion effect” (subsidies hinder or are ineffective) might prevail.

Second, it extends these theories to the context of public environmental attention, showing how normative pressures from society amplify the effects of regulative support. This addresses calls in the literature for greater attention to the interplay between formal and informal institutions in shaping corporate environmental behavior [13,14].

Third, it identifies CSR as a mediating mechanism, revealing how institutional support translates into innovation through enhanced stakeholder relationships and resource access. This extends stakeholder theory by showing how CSR serves as both an outcome of institutional pressures and a driver of innovation performance.

6.4. Limitations and Future Research

This study has several limitations that suggest directions for future research. First, the measure of public environmental attention using Baidu search indices, while validated in prior research, has limitations. Search behavior may reflect transient rather than sustained attention, and regional differences in internet penetration may affect comparability. Future research could complement search indices with other measures such as social media sentiment analysis or media coverage intensity.

Second, the CSR measure from Hexun.com, while widely used, may not capture all dimensions of social responsibility relevant to environmental protection enterprises. Future research could develop industry-specific CSR measures that better reflect the unique characteristics of environmental protection firms.

Third, while this study establishes correlations and uses instrumental variables to address endogeneity, causal identification remains challenging. Future research could exploit policy experiments or natural experiments, such as exogenous changes in subsidy programs or environmental regulations.

Fourth, the sample is limited to Chinese listed firms, which may not represent unlisted or smaller environmental protection enterprises. While our findings contribute to the global discourse on sustainable innovation by providing evidence from the world’s largest emerging economy, future research could extend the analysis to private firms and other institutional contexts—both developed and developing—to assess the generalizability of our integrated theoretical framework.

Fifth, this study focuses on one mechanism (CSR) through which subsidies affect innovation. Future research could examine other mechanisms, such as risk-taking, talent acquisition, or collaboration networks, to provide a more complete picture of how subsidies influence innovation.

7. Conclusion and Recommendations

Based on a sample of 121 listed Chinese environmental protection enterprises from 2016 to 2025, this paper empirically examines the impact of government subsidies on their sustainable innovation performance, and deeply explores the moderating role of public environmental attention and the mediating mechanism of corporate social responsibility. This study contributes to the literature by: (1) providing industry-specific evidence on the “quality and quantity” effects of subsidies, (2) demonstrating the crucial moderating role of market-driven public attention, thereby integrating government and market perspectives, and (3) revealing CSR as a key mediating mechanism, thus offering an integrated theoretical explanation grounded in institutional and signaling theories. The findings are as follows:

First, government subsidies have a significant effect of “improving both quality and quantity” on the innovation performance of environmental protection enterprises, effectively increasing both innovation output quantity and quality.

Second, public environmental attention, as an important market-driven force, positively moderates the relationship between government subsidies and innovation performance, strengthening their incentive effect. This moderating effect is stronger for innovation quality than for innovation quantity, suggesting that public scrutiny particularly encourages substantive innovation.

Third, the innovation incentive effect of government subsidies exhibits heterogeneity. Its promoting effect on innovation output quantity shows no significant heterogeneity by ownership but displays regional heterogeneity, being effective only in the eastern and western regions. The insignificance in central regions may reflect a “middle-income trap” in innovation policy. Its promoting effect on innovation output quality shows both ownership and regional heterogeneity, with more prominent incentive effects in state-owned enterprises and the western region.

Fourth, mechanism tests reveal that government subsidies partially promote the improvement of enterprise innovation performance indirectly through the path of promoting environmental protection enterprises to fulfill their social responsibilities. CSR explains approximately 2–3% of the total effect of subsidies on innovation, indicating partial mediation.

Based on the above conclusions, this paper proposes the following policy optimization recommendations:

Firstly, optimize the government subsidy mechanism to enhance fund usage efficiency. Continuous and increased fiscal support for the environmental protection industry should be maintained, expanding subsidy coverage to effectively alleviate financing constraints faced by enterprises. More importantly, a full-cycle, refined subsidy supervision and evaluation system should be established, considering the industry’s characteristics like “dual externalities,” to ensure subsidy funds are accurately directed towards R&D activities with substantive and high social value, preventing “strategic innovation,” and promoting the effective transformation of innovation outcomes [25]. This recommendation follows directly from the finding that subsidies promote both quantity and quality, but that the quality effect requires institutional oversight.

Secondly, strengthen the role of market drivers to build a synergistic “government-market” incentive framework. The government should actively use new media and other means to enhance environmental publicity and raise public environmental concern. Simultaneously, channels and mechanisms for public participation in environmental governance should be established and improved to transform public concern into market demand and supervisory pressure for corporate green innovation. This forms a synergistic effect between government subsidies and market pull, jointly stimulating corporate innovation motivation. This recommendation is grounded in the moderating effect finding—public attention amplifies subsidy effectiveness, particularly for innovation quality.

Thirdly, implement differentiated and targeted subsidy strategies. The “one-size-fits-all” subsidy model should be abandoned, fully considering heterogeneity in enterprise ownership and region. It is recommended to moderately tilt resources towards state-owned environmental protection enterprises with greater potential for improving innovation quality and towards the western region as a technological lowland, to maximize fiscal fund efficiency [26]. Concurrently, performance-based reward and punishment mechanisms should be refined, rewarding enterprises with significant innovation results and severely penalizing behaviors involving subsidy misappropriation or abuse. These differentiated strategies should be implemented gradually, with pilot programs in target regions before broader rollout, and with clear performance metrics to evaluate effectiveness.

Fourthly, guide and urge enterprises to fulfill social responsibilities, clearing the innovation mediation path. The government should promote corporate social responsibility construction through a “combination of rigidity and flexibility.” On the “rigid” side, the legalization process related to CSR should be accelerated, clarifying requirements for innovation and environmental responsibility fulfillment by environmental protection enterprises. On the “flexible” side, incentive policies such as tax benefits and green credit can be used to encourage firms to proactively disclose social responsibility reports [27]. Furthermore, a third-party evaluation and public opinion supervision system should be established and improved, setting up exemplary models and exposing non-compliant behaviors, fostering a virtuous cycle where innovation fulfills social responsibility and social responsibility drives innovation. This recommendation directly addresses the mediation mechanism identified in this study:subsidies work partly through CSR, so policies that strengthen CSR will enhance subsidy effectiveness.

In conclusion, this study demonstrates that government subsidies can effectively promote sustainable innovation in environmental protection enterprises, particularly when combined with public environmental attention and channeled through corporate social responsibility. These findings have important implications for policy design not only in China but also in other countries seeking to promote environmental innovation as a key driver of the 2030 Agenda for Sustainable Development. By quantifying the impact of policy tools and market forces, and by providing an integrated theoretical framework, this research contributes to the measurement, monitoring, and application of sustainability principles within a critical industrial sector.