Environmental Protection Tax and Green Innovation: The Mediating Role of Digitalization and ESG

Abstract

In the wave of the digital economy and “carbon neutrality”, digital governance and green governance are effective measures for firms to achieve sustainable development goals. The purpose of this paper is to examine the impact of environmental protection tax (EPT) policies on green innovation (GI) using panel data from Chinese A-share listed firms from 2010 to 2022. Using fixed effects models, we find that EPT significantly promotes firms’ GI. Mechanism tests reveal that digitalization and environment, social, and governance (ESG) performance both mediate the relationship between EPT and GI. Further analysis shows that government subsidies and analyst coverage both strengthen the effect of EPT on GI, while supplier concentration exerts a reverse moderating influence. Economic outcome tests confirm the multiple impacts of EPT in improving total factor productivity (TFP) and economic performance through GI. Our findings contribute to fulfilling the extant literature gap on the relationship between EPT and GI, and provide practical insights for firms engaged in digital and green governance.

1. Introduction

Currently, the exacerbation of the environmental pollution challenge is prompting global focus on this pressing issues. China’s Environmental Performance Index highlights a decline from 37.3 in 2020 to 28.4 in 2022, implying an increase in environmental pressures. The Chinese government faces profound challenges in reconciling economic development with environmental sustainability, as well as setting strategic goals such as “Peak Carbon by 2030” and “Carbon Neutrality by 2060” to tackle this contradiction [1]. Green innovation (GI) can be defined as a process of improving production and technologies in order to navigate environmental issues [2,3,4]. The benefits of GI tend to be accompanied by higher environmental costs, resulting in a lack of incentives for firms to engage in GI activities [5]. Increasingly, the broader literature suggests that environmental regulation is becoming a significant external factor in incentivizing and guiding firms to proactively undertake GI [6]. The historical trajectory of environmental regulation in China extends back to 1979, marked by prolonged experimentation with the sewage fee system, culminating in the formal implementation of China’s first environmental protection tax (EPT) law in 2018. The shift from a fee-based system to a tax framework consolidates the legal regulatory status of EPT, enhances the efficacy of environmental regulation, and affects the realization of firms’ green development objectives [7]. EPT levies a higher proportion of firms with higher pollutant emissions, compelling highly polluting firms to undergo their green transformation. Concurrently, the widespread acceptance of digitalization and environmental, social, and governance (ESG) performance has accelerated the incorporation of innovation strategies and governance approaches to address environmental issues. Corporate GI is the most effective avenue to promote green transition under the national “dual carbon” strategy; however, existing studies focus less on analyzing GI directly from EPT. In this context, the inquiry of this paper revolves around the pivotal question of whether EPT catalyzes firms to promote GI, and it delves into the potential mediating influence of digital governance and ESG performance. This research assumes important theoretical and practical significance for the green development of firms.

The academic discussion on EPT has garnered considerable attention, with the existing literature mainly focusing on its influence on environmental and economic performance [8,9]; however, research on the role of EPT in promoting GI remains a point of divergence in the academic field. Based on Porter’s hypothesis, some scholars assume that the innovation effect of EPT will increase firms’ research and development (R&D) investment and catalyze their GI [10]. Conversely, an opposing view contends that the cost effect of EPT will escalate firm costs and hinder the development of GI [11]. Furthermore, the interaction between innovation and cost effects is considered to yield non-linear outcomes, leading to different scenarios [12]. Apart from the above, it is worth noting that there remains a significant dearth of research revealing the mechanism of EPT in promoting GI. COVID-19 has exacerbated volatility in economic markets, exacerbating the cost and regulatory pressures of EPT, thereby impelling firms to accelerate their digital transformation initiatives [13]. This pressure, in turn, provides technological impetus to utilize digital capabilities in order to support environmental governance endeavors and concurrently enhance corporate GI [14]. Corporate EPT is shown to not only promote environmental performance but also to enhance overall corporate ESG performance [15]. Moreover, EPT optimizes strategic decisions that align with the interests of the firm and society, thereby improving innovation performance for sustainable development [16].

This empirical study uses fixed effects (FE) models to systematically analyze the impact and mechanism of EPT on GI. It uses data from Chinese A-share listed firms in Shanghai and Shenzhen from 2010 to 2022 to answer the following research questions: Firstly, does EPT improve the GI performance of listed firms in China? Secondly, does EPT influence the effect on GI through two pivotal mediating channels, namely digitalization and ESG performance? Thirdly, scrutinizing from a stakeholder perspective, to what extent do government subsidies, analyst coverage, and supplier concentration modulate the promotional effect of EPT on GI? Lastly, delving deeper into the economic consequences, does EPT affect firms’ total factor productivity (TFP) and economic performance through GI?

The possible contributions of this paper are as follows: Firstly, this paper extends the research on the economic impact of EPT and enriches the broader literature on EPT at the firm level. The EPT law has been implemented for a relatively short period, and many studies have considered the impact of environmental regulations on GI, giving less attention to their impact on the micro-level economic variables of firms. Drawing from the specific context of Chinese firms, this paper systematically examines the impact of EPT on firms’ GI and broader economic consequences. Consequently, the findings contribute to the exploration of the impact of EPT on innovation and economic incentives in developing countries, as well as expanding the literature on the “double dividend” of EPT. Secondly, this paper discovers the intricate mediating mechanism of digitalization and ESG performance, which supplements the emerging literature in both fields [17]. The study finds that firms can navigate the regulatory effects of EPT through digital and green governance mechanisms, thereby augmenting the existing body of knowledge on the green governance of EPT. The findings enhance scholarly insights while offering firms pragmatic guidance on how to strategically respond to environmental policies. Thirdly, by combining firm behavior, stakeholder behavior, and GI, this paper explores the moderating effect of interventions from different stakeholders on the effects of EPT on GI. This paper innovatively analyzes the boundary conditions for EPT to play the role of innovation incentives from an integrated stakeholder perspective. The findings explain the moderating effects stemming from government subsidies, analyst coverage, and supplier concentration on firms’ decisions to engage in GI. The study not only advances theoretical understanding but also offers practical suggestions for the application and design of EPT alongside considerations of stakeholder behavior.

The remaining sections of this paper are organized as follows: Section 2 offers a literature review and research hypotheses; Section 3 describes the research methodology, encompassing models, variables, and data; Section 4 conducts an in-depth analysis of empirical results, including benchmark regressions, mediated effects regressions, and robustness tests; Section 5 further analyzes the moderating effects of government subsidies and analyst coverage from a stakeholder perspective, as well as the analysis of economic consequences in terms of total factor productivity and economic performance; and Section 6 provides the main conclusions and implications of the paper.

2. Literature Review and Research Hypotheses

2.1. The Impact of EPT and GI

EPT operates as a market incentive-based environmental regulation tool, strengthening law enforcement and environmental constraints on firms, thereby increasing the costs of illegal activities by firms [18]. The “double dividend” effect of EPT, including environmental and economic dividends, remains a hot topic of academic discussion. In terms of environmental dividends, EPT has a reduction effect on pollution emissions [19,20]. Given the development of a national strategy for the 2030 carbon-neutral target, the impact of EPT on carbon emissions has received more specialized research. Scholars assume that, in both the short and long term, EPT can reduce carbon intensity, minimize carbon emissions, and foster environmental innovation [21]. Besides, EPT contributes to enhancing urban pollution control and promoting environmental performance [22]. The impact of increased EPT on improving environmental performance in heavily polluting firms further indicates the significant environmental dividend effect of EPT [23]. Turning to economic dividends, the realization of the “double dividend” role of EPT is obvious in China, Finland, and Malaysia [24]. The “double dividend” effect of EPT has temporal heterogeneity, which is more pronounced in stimulating long-run economic growth and reducing short-term pollution [25]. Notably, the implementation of the EPT law has led to an increase in the production costs of highly polluting firms, constricting investment in production and operations, and reducing the labor share of firms [26]. However, the above research on the impact of EPT focuses more on the meso-level and macro-level, with relatively few studies undertaken at the micro-firm level.

Does the “double dividend” effect of EPT extend to the promotion of GI at the micro-level? Current research at the micro-level focuses more on heavily polluting firms and manufacturing industries, with limited exploration of the broader industry-wide influence of EPT. Since the official implementation of the EPT law in 2018, scholars have increasingly turned to conducting analyses of policy applications in order to study the economic role of EPT, while fewer studies have used objective firm-level data for overall analyses. Therefore, this paper attempts to analyze whether the impact of EPT on GI generates a dividend effect from the perspective of listed firms across the industry, using enterprise EPT data. At present, scholars exhibit varied perspectives on the relationship between EPT and GI. Firstly, regarding the incentive effect, the Porter hypothesis posits that environmental regulations act as stimuli for technological innovation and competitiveness, with the innovation effects partially offsetting the cost effect induced by environmental regulations [27]. Deng et al. conclude that the imposition of EPT intensifies legitimacy pressure on firms, forcing polluting entities to improve legitimacy management efforts and increase R&D investment [28]. Zhao et al. confirm that EPT improves the green R&D efficiency of firms, particularly enhancing the GI of heavily polluting firms [18]. Secondly, regarding the inhibiting effect, the imposition of EPT increases the environmental costs borne by firms, suppresses the input of innovation resources, and works against the GI of firms. As the EPT rate escalates, the environmental regulatory effects intensify, substantially increasing firms’ production costs [29]. Wang and Yu claim that the current EPT rate of China’s resource-based industries is relatively low and environmental externalities cannot be fully internalized, which inhibits GI [30]. Additionally, Du et al. argue that the increase in environmental protection costs due to EPT accelerates capital renewal, but concurrently depresses the quantity and quality of GI [31]. Lastly, in considering the non-linear relationship between EPT and GI, there are also various opinions among scholars. Some scholars assume that there is a U-shaped relationship between EPT and GI. In the short term, Jiang et al. argue that EPT may initially suppress GI, but could increase it in the long run, due to lagging tax effects [12]. Conversely, an alternative perspective posits an inverted U-shaped relationship between EPT and GI. Wei et al. confirm that China’s current EPT has not yet reached the inflection point of the inverted U-shaped curve, indicating that EPT is still at the stage of promoting GI [32].

Currently, China’s EPT law has increased the intensity of enforcement. EPT is essentially a “polluter pays” tax, thus Zheng et al. argue that EPT increases environmental costs and indirectly induces firms to engage in GI [33]. The regulatory impact of EPT is manifested as imposing environmental pressure on taxable pollutants, forcing firms to address the internalization of negative environmental externalities; therefore, firms are encouraged to undertake GI, reduce pollution emissions, and promote efforts in emissions reduction and pollution control. Berrone et al. conclude that the imposition of EPT raises firms’ compliance costs, prompting adaptive responses to regulatory pressures, such as improving production lines to achieve green production [34]. Furthermore, Tchorzewska et al. support the hypothesis that the increasing of EPT can promote the adoption of green technologies by firms and improve the level of green investment [35]. The long-term “dual carbon” strategy demonstrates China’s commitment to green transformation, accompanied by substantial governmental support for the green development of industries and firms. In the face of legitimacy pressure arising from environmental regulations, Su et al. conclude that firms tend to make decisions in areas with strong governmental support, proactively engaging in green transformation initiatives [36]. Drawing on Porter’s innovation effect, this paper posits that EPT can serve to promote GI at the current stage. In the short term, the rising environmental costs prompt firms to actively adjust green production strategies, intensify green inputs, improve the efforts to promote GI, and enhance environmental performance. Over the long term, EPT serves to enhance economic sustainability, improve environmental quality, and strengthen the competitive advantages of firms [37]. Based on the above analysis, we propose Hypothesis 1:

Hypothesis 1 (H1). 

EPT exerts a promoting effect on the GI of firms.

2.2. The Mediating Role of Digitalization

Due to the enduring impact of COVID-19 on the real economy, digitalization has developed rapidly in China’s firms, emerging as a prevalent model of corporate governance. In 2016, the official promulgation of China’s EPT law clarified the conversion of firms’ pollution discharge fees into tax regulations with stronger enforcement. Thus, firms have had to face tougher environmental regulatory pressure. Market turbulence and uncertainty act as both a challenge and a motivation for firms, forcing them to improve digital governance for competition in the new era [38]. Currently, scholars have found the positive impact of digitalization on the environmental aspects of firms, noting its contributions to GI [39]; however, there are few studies on environmental tax regulation and digitalization. The positive impact of digitalization on GI may encourage firms to improve digital performance in their production operations, thereby indirectly affecting the effects of EPT on GI. Firstly, EPT prompts firms to improve managers’ awareness of environmental governance, construct a green brand image, and enhance corporate reputation. Zhou et al. argue that the application of digital technologies, such as big data and cloud computing, provides technical support for firms to establish environmentally sustainable production processes [40]. Chen et al. argue that the environmental reputation pressures induced by EPT can incentivize digital transformation, thereby enhancing the green competitive advantages of firms [41]. Secondly, EPT increases the cost burden on firms, and creates additional capital requirements. EPT-induced cost pressures prompt firms to seek more efficient strategies for pollution and emission reduction, develop differentiated products to alleviate cost pressures, and reduce environmental costs. Chen and Zhang claim that digitalization helps firms to provide timely feedback on the needs of customers and consumers, enabling customized production and improving the performance of smart production and innovation [42]. Thirdly, information about environmental regulation and GI often exhibits a lag, and the effectiveness of environmental governance cannot be seen in the short term [43]. Li et al. argue that the information-gathering capabilities of digital governance help to realize the interconnection of information flows [44]. Moreover, Gupta and Bose conclude that digital transformation enables firms to rapidly respond to environmental issues, improve green decision-making efficiency, and enhance environmental performance [45]. In summary, EPT serves to influence GI through digitalization. Based on the above analysis, we propose Hypothesis 2:

Hypothesis 2 (H2). 

Digitalization plays a mediating role in the relationship between EPT and GI.

2.3. The Mediating Role of ESG Performance

ESG performance has emerged as a key indicator for measuring sustainable development, generating significant stakeholder interest and prompting firms to focus on the performance and disclosure practices of ESG performance. Currently, scholars mainly study the relationship between ESG performance and enterprise behavior, but there are few studies specifically examining the impact of ESG performance and GI. ESG performance exerts a promoting effect on firm innovation and performance [16]. The information disclosure of ESG performance reduces information asymmetry, diminishes credit risk, and alleviates financing constraints [46]. The impact of ESG performance on GI prompts firms with better ESG performance to focus more on environmental protection, thus indirectly affecting the role of EPT in GI. Firstly, concerning environmental performance, the imposition of EPT helps firms to enhance green awareness, improve environmental performance, promote environmental reputation, and obtain increased recognition and support from a broad range of stakeholders [47]. Building upon the resource-based theory, Hart and Dowell claim that the active disclosure of ESG information, encompassing comprehensive performance of environmental protection, social responsibility, and corporate governance, facilitates the establishment of a good image and cultivates a reputation for firms’ responsibility [48]. To maintain a positive reputation and the recognition brought by environmental performance, firms may allocate additional resources to GI activities, turning from a reactive to a proactive focus on GI [28]. Secondly, from the perspective of social performance, the imposition of EPT helps firms establish a positive image with a robust sense of social responsibility. Xie et al. argue that firms with a positive green image enhance enterprise self-confidence and improve stakeholders’ trust, thus more easily improving economic return on green product innovation [49]. Based on the signaling theory, El et al. claim that good ESG performance guides firms to positively undertake environmental responsibilities, attracting green investment, reducing financing costs, and exerting a beneficial impact on the GI activities of firms [50]. Thirdly, in terms of corporate governance performance, the imposition of EPT subjects firms to tax pressures, forcing managers to focus on national tax policies and fulfill environmental responsibilities [51]. Owing to the increased managerial attention to environmental laws, firms pay more attention to maintaining organizational legitimacy and sustainable strategies. This green concern manifests in the enhancement of green technology and an intensified focus on green R&D innovation in order to achieve long-term competitive advantage and value growth. In summary, EPT serves to influence GI through ESG performance. Based on the above analysis, we propose Hypothesis 3:

Hypothesis 3 (H3). 

ESG performance plays a mediating role in the relationship between EPT and GI.

The conceptual model in this research is shown in Figure 1.

3. Model, Variable, and Data

3.1. Model Specification

Based on the research of [52], this paper constructs FE models and mediating effect models to examine the impact of EPT on GI. Formula (1) tests the direct relationship between EPT and GI; Formulas (1)–(3) test the mediating effect role of digitalization; and Formulas (1), (4) and (5) test the mediating role of ESG performance through a step-by-step approach.

$${\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\beta }}_2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(1)

$${\mathrm{D}\mathrm{I}\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\alpha }}_{0\mathrm{a}}+{\mathsf{\alpha }}_{1\mathrm{a}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\alpha }}_{2\mathrm{a}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(2)

$${\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\gamma }}_{0\mathrm{a}}+{\mathsf{\gamma }}_{1\mathrm{a}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\gamma }}_{2\mathrm{a}}{\mathrm{D}\mathrm{I}\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\gamma }}_{3\mathrm{a}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(3)

$${\mathrm{E}\mathrm{S}\mathrm{G}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\alpha }}_{0\mathrm{b}}+{\mathsf{\alpha }}_{1\mathrm{b}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\alpha }}_{2\mathrm{b}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(4)

$${\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\gamma }}_{0\mathrm{b}}+{\mathsf{\gamma }}_{1\mathrm{b}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\gamma }}_{2\mathrm{b}}{\mathrm{E}\mathrm{S}\mathrm{G}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\gamma }}_{3\mathrm{b}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(5)

The subscripts “i”, “s”, “t” and “p” denote firm, industry, year, and province, respectively. GI represents the level of GI. EPT represents environmental protection-related taxes. DIGI represents digitalization as a mediator variable. ESG represents ESG performance as another mediator variable. Controls represent a matrix of control variables for firms, including firm size (SIZE), asset-liability ratio (LEV), firm growth (GROWTH), firm value (TQ), cash hold (CASH), executive shareholding (EHOLD), return on assets (ROA), CEO duality (DUAL), and state ownership (SOE). Additionally, industry (Ind), year (Year), and province (Prov) fixed effects are controlled. Given the inclusion of these triple fixed effects, using simple firm-level residues may underestimate the standard error; thus, ε in our regression model represents a random error term clustered at the industry-year-province level. Moreover, β, α, and γ denote regression coefficients. In the mediator effect model, we mainly focus on the coefficients of β₁, α_1a, α_1b, γ_1a, γ_1b, γ_2a, γ_2b, α_1aγ_2a, and α_1bγ_2b. Specifically, β₁ represents the total effect of EPT on GI; α₁ represents the effect of EPT on mediator variables; γ_1a and γ_1b represent the direct effect of EPT on GI; γ_2a represents the mediating effect of digitalization on GI after EPT is under control; and γ_2b represents the mediating effect of ESG performance on GI after EPT is under control. α_1aγ_2a indicates the indirect effect of EPT transmitted through digitalization, and α_1bγ_2b indicates the indirect effect of EPT transmitted through ESG performance.

3.2. Variable Selection and Interpretation

3.2.1. Dependent Variable: Green Innovation (GI)

As there is no separate disclosure for green R&D, scholars have widely used the applications and grant numbers of green patents as proxy variables to measure GI [53]. The number of green patent applications can reflect the development of firms’ GI level more accurately than the number of authorizations; therefore, following [54], this paper uses “the logarithm of (the number of green invention patent applications of listed firms + 1)” to measure GI.

3.2.2. Independent Variable: Environmental Protection Tax (EPT)

Following [55], we hold the generalized environmental protection tax (EPT) refers to the aggregate of taxes and fees related to environmental protection. Specifically, the total environmental protection tax consists of the resource tax, urban land use tax, urban maintenance and construction tax, vehicle purchase tax, farmland occupation tax, vehicle and vessel tax, pollution discharge fee, environmental protection tax, and other taxes associated with environmental protection. In this paper, we use “the logarithm of (the total environmental protection tax + 1)” to measure EPT.

3.2.3. Mediating Variables

(1)

Digitalization (DIGI). Digitalization refers to the application degree of firms’ digital technology. Following [39], the word frequency related to digital technology, such as “artificial intelligence”, “big data”, “blockchain”, and “cloud computing” reported in firms’ annual reports, calculated by using the text mining method, can reflect the importance of firms to digital strategy. Therefore, this paper selects “the logarithm of (related digital technology word frequency in the annual report +1)” to measure digitalization (DIGI).

(2)

ESG performance (ESG). Following [56], this paper uses “ESG indicators in the ESG rating system of China’s Sino-Securities Index” as a proxy variable for ESG performance. ESG performance is categorized into nine grades: AAA, AA, A, BBB, B, B, CC, CC, and C. Then, ESG ratings are converted into a numerical scale ranging from 1 to 9, corresponding to CCC through AAA grades.

3.2.4. Control Variables

Following [57], to eliminate potential interference with estimation results, we use the following firm-level control variables: Firm size (Size), denoted as “the logarithm of (the total assets + 1)”; asset liability ratio (LEV), denoted as the ratio of “total liability/total assets”; firm growth (GROWTH), denoted as “current year’s sales revenue/previous year’s sales revenue − 1”; firm value (TQ), denoted as the ratio of “market value/total assets”; cash hold (CASH), denoted as the ratio of “monetary capital holdings/total assets”; executive shareholding (EHOLD), denoted as the ratio of “number of shares held by executives/total number of shares”; return on assets (ROA), denoted as the ratio of “net profit/total assets”; CEO duality (DUAL), denoted as a dummy variable that “if the board director and CEO are the same person equals 1, otherwise 0”; and state ownership (SOE), denoted as a dummy variable that “state-owned firms equal 1, otherwise 0”. In addition, the FE model incorporates an industry dummy (Ind), year dummy (Year), and province dummy (Prov).

The specific definitions of the relevant variables are shown in

Table 1. Definitions of key variables.

Variable	Symbol	Measurement of Variable	Reference
Dependent Variable
Green innovation	GI	The logarithm of (the number of green invention patent applications + 1)	[54]
Independent Variable
Environmental protection tax	EPT	The logarithm of (the total environmental protection tax + 1)	[55]
Mediator Variable
Digitalization	DIGI	The logarithm of (related digital technology word frequency in the annual report +1)	[39]
ESG performance	ESG	ESG indicators from the ESG rating system of China’s Sino-Securities Index	[56]
Control Variable
Firm size	SIZE	The logarithm of (the total assets+1)	[57]
Asset liability ratio	LEV	Total liability/Total assets	[57]
Firm growth	GROWTH	Current year’s sales revenue/Previous year’s sales revenue − 1	[57]
Firm value	TQ	Market value/Total assets	[10]
Cash hold	CASH	Monetary capital holdings/Total assets	[58]
Executive shareholding	EHOLD	Number of shares held by executives/Total number of shares	[28]
Return on assets	ROA	Net profit/Total assets	[59]
CEO duality	DUAL	If the board director and CEO are the same person equals 1, otherwise 0	[57]
State ownership	SOE	State-owned firms equal 1, otherwise 0	[57]

.

3.3. Data Collection

This paper selects A-share listed firms in Shanghai and Shenzhen as the research subjects of investigation. The sample period ranges from 2010 to 2022; this period was chosen to accord with the increasing availability of firms’ ESG annual reports since 2010. Green patent data are collated by the matching of data from the State Intellectual Property Office with the WIPO’s IPC Green Inventory. ESG data are processed from the official website of the Sino-Securities Index in China, including specific rating scores and rating data ranging from AAA to CCC grades. All other data are selected from the CSMAR and WIND databases, along with the firms’ annual reports. The sample data are filtered to exclude firms in the financial industry and those listed as ST or PT or delisted from the stock market, as well as missing core data. The main continuous variables are winsorized at 1% and 99% to avoid the influence of extreme values. Finally, we have 33,220 samples from 4468 listed firms, and Stata16.0 is employed for empirical tests. Figure 2 shows the distribution of industry and year of sample firms.

4. Empirical Results and Analysis

4.1. Descriptive Statistics and Collinearity Analysis

Table 2. Descriptive statistics and collinearity test.

Variable	Observation	Mean	Std. Dev.	Minimum	Maximum	VIF
GI	33,220	0.283	0.663	0	3.638
EPT	33,220	15.471	1.567	8.591	20.311	2.85
DIGI	33,220	1.403	1.399	0	5.257	1.08
ESG	33,220	4.145	1.063	1	8	1.15
SIZE	33,220	22.167	1.280	19.478	26.456	3.62
LEV	33,220	0.415	0.207	0.028	0.930	1.87
GROWTH	33,220	0.171	0.406	−0.673	4.474	1.09
TQ	33,220	2.039	1.371	0.799	17.676	1.21
CASH	33,220	0.168	0.132	0.007	0.825	1.29
EHOLD	33,220	0.149	0.203	0	0.705	1.5
ROA	33,220	0.037	0.067	−0.642	0.223	1.44
DUAL	33,220	0.305	0.460	0	1	1.13
SOE	33,220	0.329	0.470	0	1	1.49
Mean VIF						1.64

presents the summary of descriptive statistics and the values of the variance inflation factor (VIF). As shown in Table 2, the average level of GI stands at 0.283, indicating the relatively low and diverse extent of GI across the sample firms. All other variables exhibit values within reasonable limits. The VIF test conducted on the main variables reveals values below 5 for each variable, and the average VIF is much less than 10. This VIF outcome indicates that there is no multicollinearity between variables.

4.2. Main Regression Results

4.2.1. Regression Results of Baseline Regression

Table 3. Regression results of baseline regression.

Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	OLS		FE				RE
	GI	GI	GI	GI	GI	GI	GI
EPT	0.089 ***	0.025 ***	0.008 *	0.017 **	0.016 ***	0.012 ***	−0.001
	(39.08)	(6.73)	(1.65)	(2.50)	(4.17)	(2.86)	(−0.20)
SIZE		0.125 ***	0.103 ***	0.144 ***	0.142 ***	0.143 ***	0.095 ***
		(24.42)	(10.37)	(11.07)	(19.90)	(19.41)	(10.22)
LEV		−0.010	−0.061 *	0.063	0.072 ***	0.093 ***	0.036
		(−0.43)	(−1.77)	(1.35)	(2.98)	(3.88)	(1.18)
GROWTH		−0.028 ***	−0.025 ***	−0.043 ***	−0.042 ***	−0.042 ***	−0.025 ***
		(−3.11)	(−4.97)	(−5.42)	(−5.31)	(−5.35)	(−4.81)
TQ		0.025 ***	0.008 ***	0.017 ***	0.017 ***	0.017 ***	0.008 ***
		(8.85)	(3.42)	(4.24)	(5.67)	(5.45)	(3.02)
CASH		0.187 ***	−0.088 ***	0.156 ***	0.146 ***	0.189 ***	0.035
		(6.25)	(−2.76)	(3.05)	(4.84)	(6.15)	(1.15)
EHOLD		0.231 ***	0.010	0.095 **	0.063 ***	0.057 ***	0.039
		(10.97)	(0.24)	(2.48)	(3.11)	(2.79)	(1.29)
ROA		−0.012	−0.041	0.219 ***	0.210 ***	0.248 ***	0.093 **
		(−0.19)	(−0.91)	(2.86)	(4.30)	(5.16)	(2.07)
DUAL		0.048 ***	−0.006	0.038 **	0.033 ***	0.031 ***	−0.001
		(5.87)	(−0.60)	(2.55)	(3.89)	(3.63)	(−0.12)
SOE		0.030 ***	−0.002	0.061 ***	0.066 ***	0.072 ***	0.048 **
		(3.29)	(−0.09)	(2.81)	(6.89)	(7.50)	(2.45)
Constant	−1.088 ***	−3.012 ***	−2.089 ***	−3.314 ***	−3.259 ***	−3.208 ***	−2.103 ***
	(−30.85)	(−36.46)	(−10.75)	(−12.47)	(−21.05)	(−20.05)	(−11.23)
Province fixed effects	No	No	No	No	Yes	Yes	Yes
Industry fixed effects	No	No	No	Yes	Yes	Yes	Yes
Year fixed effects	No	No	No	No	No	Yes	Yes
Observations	33,220	33,220	33,220	33,220	33,220	33,220	33,220
Adjust R2	0.0439	0.0657	0.0239	0.153	0.160	0.163

Notes: ***, **, * denote significance at the 1%, 5%, and 10% level, respectively. T-statistics are reported in parentheses.

empirically tests Hypothesis 1, examining the direct effect of EPT on GI. Columns (1) to (2) show the consistently positive and statistically significant coefficient for EPT in the ordinary least squares (OLS) model, regardless of the inclusion of control variables. With the introduction of individual fixed effects and heteroscedastic robust standard error, the coefficient in column (3) remains significantly positive at the 10% level. Similarly, considering industry fixed effects and clustering standard errors at the firm level, the coefficient in column (4) remains significantly positive at the 5% level. Further, considering industry-province fixed effects and clustering standard errors at the industry-year-province level, the coefficient in column (5) becomes significantly positive at the 1% level. Even considering industry-year-province fixed effects and clustering standard errors at the industry-year-province level, the coefficient in column (6) remains significantly positive at the 1% level. However, when considering industry-year-province fixed effects and clustering standard errors at the firm level using the random effects (RE) model, the coefficient in column (7) is no longer significant. The effectiveness of the FE model is verified by an overidentifying restriction test on the RE regression. Through diverse regression methods, controlling for different fixed effects levels, and clustering at various levels, the incentive effect of EPT on firms’ GI is consistently empirically supported. Thus, Hypothesis 1 (H1) is supported. Specifically, EPT is found to increase firms’ GI by 6.64% (0.012 × 1.567/0.283). This finding is consistent with [60], reinforcing the notion of the incentive effect of environmental taxes on GI, in light of an increase of 9.16% (0.429 × 0.157/0.735) in the number of green invention patents and 6.09% (0.429 × 0.151/1.063) in the number of green patents.

4.2.2. Regression Results of the Mediating Effect

In this section, we explore two potential influence mechanisms: “EPT—digitalization—GI” and “EPT—ESG performance—GI”.

(1)

Mediating mechanism test of digitalization. Panel A of

Table 4. Regression results of the mediating effects test.

Panel A: Digitalization
Variable	(1)	(2)	(3)	(4)	(5)
	GI	DIGI	DIGI	GI	GI
EPT	0.012 ***	0.027 ***	0.020 ***	0.010 **	0.010 **
	(2.86)	(3.45)	(3.05)	(2.43)	(2.53)
DIGI				0.064 ***	0.062 ***
				(15.73)	(13.28)
Constant	−3.208 ***	−1.998 ***	−1.177 ***	−3.081 ***	−3.135 ***
	(−20.05)	(−12.03)	(−7.74)	(−19.94)	(−19.99)
Control variables 1	Control	Control	Control	Control	Control
Province fixed effects	Yes	Yes	Yes	Yes	Yes
Industry fixed effects	Yes	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes	Yes
Observations	33,220	33,220	33,220	33,220	33,220
Adjust R2	0.163	0.406	0.378	0.174	0.171
Panel B: ESG Performance
Variable	(6)	(7)	(8)	(9)	(10)
	GI	ESG	ESG	GI	GI
EPT	0.012 ***	0.051 ***	0.003 ***	0.008 **	0.008 **
	(2.86)	(7.12)	(7.02)	(2.12)	(2.12)
ESG				0.060 ***	0.932 ***
				(15.83)	(16.30)
Constant	−3.208 ***	−0.871 ***	3.971 ***	−3.156 ***	−6.910 ***
	(−20.05)	(−5.38)	(368.91)	(−19.88)	(−22.07)
Control variables	Control	Control	Control	Control	Control
Province fixed effects	Yes	Yes	Yes	Yes	Yes
Industry fixed effects	Yes	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes	Yes
Observations	33,220	33,220	33,220	33,220	33,220
Adjust R2	0.163	0.180	0.194	0.171	0.171

Notes: *** and ** denote significance at the 1% and 5% level, respectively. T-statistics clustered at the industry-year-province level are reported in parentheses. Standard errors are clustered at the industry-year-province level. ¹ For space constraints, the control variables are not reported in the detailed results; the same applies to the table below.

empirically tests Hypothesis 2, examining whether digitalization serves as a mediating mechanism between EPT and GI. We use a step-by-step mediation effect model for regression. The first-step results of the second step are shown in column (1), and these align with the benchmark regression results. The results of the second step are shown in column (2), where the EPT coefficient is significant at the 1% level, indicating that EPT improves firms’ digitalization level (DIGI). The results are consistent with the results of [13]. The results of the third step are shown in column (4), where the EPT coefficient is significant at the 5% level and the DIGI coefficient is significant at the 1% level. These findings indicate that digitalization partially mediates the effect of EPT on GI; thus, Hypothesis 2 (H2) is supported. The results reveal that appropriate EPT improves firms’ requirements for information transparency and sharing, prompting the application of digital technology in production and operations. EPT has a driving effect on digital technology, and the improvement of digitalization promotes firms’ GI, demonstrating the significance of the “digital effect” of EPT. Furthermore, “the logarithm of (related digital technology word frequency in the annual management discussion + 1)” is selected as the substitute variable for DIGI. The results of the robustness regression are shown in columns (3) and (5), affirming the robustness of the mediating mechanism of digitalization.

(2)

Mediating mechanism test of ESG performance. Panel B of Table 4 empirically tests Hypothesis 3, examining whether ESG performance serves as a mediator mechanism between EPT and GI. The first-step results of the second step are shown in column (6), and these align with the benchmark regression results. The results of the second and third steps of the ESG’s mediation regression model are shown in columns (7) and (9). The coefficients of EPT and ESG performance are significant, indicating that ESG performance has a partial mediating role in the effect of EPT on GI; thus, Hypothesis 3 (H3) is supported. The results are consistent with the findings of [51]. The results reveal that the tax pressure from appropriate EPT increases firms’ demand for comprehensive environmental management, forcing firms to reduce pollutant emissions and undertake green production transformations. EPT has a driving effect on environmental governance and green development, and the improvement of ESG performance promotes firms’ GI, demonstrating the significance of the “green effect” of EPT. Furthermore, “the logarithm of (the specific rating score of ESG on the official website of the Sino-Securities Index in China + 1)” is selected as the substitute variable for ESG performance. The results of robustness regression are shown in columns (8) and (10), affirming the robustness of the mediating mechanism of ESG performance.

(3)

Sobel Test and Bootstrap Test. To further test the mediation effect, the Sobel test and Bootstrap test are performed on the mediating effects of digitalization and ESG performance. The indirect effects of the tests can be seen in

Table 5. The indirect effect of Sobel and Bootstrap tests.

	Sobel Test				Bootstrap Test
Variable	Coef.	Std. Err.	Sobel Z	Med. Prop.	Coef.	Std. Err.	[95% Conf. Interval]
DIGI	0.002	0.001	3.370 ***	15.00%	0.0036	0.0005	[0.0026, 0.0046]
ESG	0.003	0.000	6.495 ***	26.70%	0.0026	0.0004	[0.0017, 0.0034]

Notes: *** denote significance at the 1% level.

. Firstly, controlling for both fixed effects and clustering at the industry-year-province level for the Sobel test, that test shows that the Z-statistic of the indirect effect of digitalization is significant at the 1% level, with a mediation effect proportion of 15%. The Z-statistic of the indirect effect of ESG performance is significant at the 1% level, with a mediation effect proportion of 26.7%. Then, the bias-corrected percentile bootstrap method is employed for sampling regression, using a sample size of 500, while controlling for industry-fixed effects. The Bootstrap test in Panel B shows that the 95% confidence intervals of digitalization and ESG performance are both above 0, indicating a significant mediation effect.

In summary, EPT promotes GI through two intermediary mechanisms; in other words, EPT can enhance GI by improving digitalization and ESG performance.

4.3. Robustness Test

To reduce possible endogenous problems, we use five tests to ensure the robustness of our conclusion: variable replacement and lag, the instrument variable (IV) method, the Heckman two-step method, alternative regression methods, and propensity score matching (PSM).

4.3.1. Replacement and Lagging of Variables

To further test the reliability of our conclusions, we conduct robustness tests by replacing variables from two perspectives, and the results are shown in

Table 6. Robustness Test I: Replacement and lagging of variables.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	GI	GI	GIO	GUMA	GUMO	L.GI
EPT	0.009 **		0.012 ***	0.013 ***	0.011 ***	0.011 **
	(2.21)		(4.25)	(3.57)	(2.91)	(2.56)
L.EPT		0.018 ***
		(4.28)
Constant	−3.224 ***	−3.227 ***	−1.737 ***	−2.199 ***	−2.214 ***	−3.176 ***
	(−20.20)	(−19.88)	(−14.26)	(−17.43)	(−16.58)	(−18.43)
Control variables	Control	Control	Control	Control	Control	Control
Province fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Industry fixed effects	Yes	Yes	No	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations	33,220	31,312	33,220	33,220	33,220	30,850
Adjust R2	0.163	0.165	0.0677	0.154	0.158	0.164

Notes: *** and ** denote significance at the 1% and 5% level, respectively. T-statistics clustered at the industry-year-province level are reported in parentheses. Standard errors are clustered at the industry-year-province level.

. The first aspect is the replacement of the independent variables, using variables lagged by one period. We regress using “EPT excluding the pollutant discharge fee” as the independent variable, and the results are shown in column (1). Subsequently, we regress using “the lagged EPT” (L.EPT), and the results are shown in column (2). The second aspect is the replacement of the dependent variables, using variables lagged by one period. Given the one-sided nature of green invention patents, we use “the logarithm of (the number of green invention patents obtained + 1)” (GIO), “the logarithm of (the number of green utility model patent applications + 1)” (GUMA), and “the logarithm of (the number of green utility model patents obtained + 1)” (GUMO) as the dependent variables in regression. The results are shown in columns (3), (4), and (5). Finally, we use “the lagged GIA” (L.GIA) to regress again, and the results are shown in column (6). In conclusion, the regression results are consistent with our baseline regression, confirming the robustness of our findings.

4.3.2. IV

To eliminate potential endogenous concerns, we further use IV regressions to examine the relationship between EPT and GI. In this section, we implement the two-stage least squares method (2SLS). The payment of EPT is influenced by firm decisions, which are easily influenced by the similar activities of other firms in the surrounding areas [18]; therefore, “the mean EPT of other firms in the same industry” (M.EPT), the same province, and in the same year is selected as the first IV for the empirical test. Furthermore, there may be a reverse causality between EPT and GI. To eliminate this possible impact, this paper further selects “the EPT lagged one period” (L.EPT) as the second IV for the empirical test. Finally, we use M.EPT and L.EPT together as a joint IV for regression. The results of these three tests are shown in

Table 7. Robustness Test II: IV-2SLS.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	IV-2SLS: First Stage			IV-2SLS: Second Stage
	EPT	EPT	EPT	GI	GI	GI
EPT				0.160 ***	0.031 ***	0.022 ***
				(4.86)	(5.96)	(4.30)
M.EPT	0.166 ***		0.014 **
	(13.72)		(2.06)
L.EPT		0.781 ***	0.789 ***
		(77.92)	(84.46)
Constant	−10.026 ***	−1.641 ***	−1.852 ***	−2.190 ***	−3.284 ***	−3.343 ***
	(−46.44)	(−15.30)	(−13.77)	(−6.99)	(−21.44)	(−21.00)
Control variables	Control	Control	Control	Control	Control	Control
Province fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Industry fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effects	No	No	Yes	No	No	Yes
Observations	33,220	31,312	31,312	33,220	31,312	31,312
Adjust R2				0.126	0.162	0.165
Kleibergen-Paap rk LM statistic				129.834 ***	774.741 ***	798.429 ***
Kleibergen-Paap rk Wald F statistic				188.208	6071.609	3567.111
				[16.38]	[16.38]	[19.93]
Hansen J statistic				0.000	0.000	0.22
Endogenous test				21.085 ***	22.245 ***	11.239 ***

Notes: ** and, ** denote significance at the 1% and 5% level, respectively. T-statistics clustered at the industry-year-province level are reported in parentheses.

and demonstrate a positive relationship between EPT and GI. From the empirical results, the first stage shows that the coefficients of IV are significant, satisfying the correlation conditions. The second stage shows the following results: the underidentification tests (Kleibergen-Paap rk LM statistic) are significant at the 1% level, rejecting the original hypothesis; the weak identification tests (Kleibergen-Paap rk Wald F statistic) are far above 10, rejecting the original hypothesis; the overidentification tests (Hansen J statistic) are not appliable for solo IV, and the P-value of joint IV in column (6) is 0.22, falling in the range of (0.1–0.25), confirming the validity of the selected IV; the endogenous tests are significant at the 1% level, rejecting the original hypothesis. In conclusion, the IVs chosen in this paper are deemed reasonable, and the results of IV-2SLS regression are reliable.

4.3.3. Heckman Two-Step Model

We use the Heckman two-step method to address the potential sample selection problem of whether firms carry out GI activities. The influence of firm EPT on GI is carried out in two steps as follows: The first step is the Probit regression of the decision of firms to undergo GI activities, with “the dummy variable of whether firms carry out GI” (GI_IF) as the dependent variable. Control variables are consistent with the benchmark regression, and year-province fixed effects are controlled. The second step is to put the inverse Mills ratio (IMR) obtained in the first step into the model for regression. The results of the Heckman two-step method are shown in columns (1) and (2) of

Table 8. Robustness Test III: Heckman, NB, ZINB, and PSM.

Variable	(1)	(2)	(3)	(4)	(5)
	Heckman Two-Step		NB	ZINB	PSM
	GI_IF	GI	GI	GI	GI
EPT	0.045 ***	0.254 ***	0.032 *	0.032 *	0.011 *
	(4.25)	(6.42)	(1.89)	(1.86)	(1.71)
IMR		7.700 ***
		(7.21)
Constant	−5.705 ***	−40.207 ***	−12.466 ***	−11.795 ***	−1.845 ***
	(−25.61)	(−7.88)	(−36.35)	(−28.07)	(−9.08)
Control variables	Control	Control	Control	Control	Control
Province fixed effects	Yes	Yes	Yes	Yes	Yes
Industry fixed effects	No	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes	No
Observations	33,220	6791	33,220	33,220	14,580
Adjust R2		0.207			0.636
Pseudo R2	0.0533		0.141
Log likelihood	−15,928	−7188	−19,151	−19,081	−4801

Notes: *** and * denote significance at the 1% and 10% level, respectively. T-statistics clustered at the industry-year-province level are reported in parentheses.

. The findings show that, even after considering the sample selection bias regarding whether firms choose to conduct GI, the EPT of firms still has a significant promotion effect on the number of green invention patents, indicating that the sample selection issue does not lead to significant estimation errors.

4.3.4. Replacement of the Regression Method

Given the discreteness of patent application numbers and the prevalence of zero values in the sample, negative binomial (NB) regression and zero-inflated negative binomial (ZINB) regression are further applied in this paper. The coefficients in columns (3) and (4) of Table 8 are significantly positive at the 10% level, again confirming that the effect of EPT on GI persists across different regression models, highlighting the robustness of our findings.

4.3.5. PSM

To eliminate the sample self-selection bias, we use the PSM method to alleviate possible endogenous questions. In line with [61], firms in heavy polluting industries are selected as the experimental group, while firms in other industries are selected as the control group. We set the “dummy variable of whether the firm is of a heavy pollution industry” (TREAT) for the PSM regression. Firstly, we use local linear regression matching to pair the samples, selecting covariates consistent with the benchmark regression. The nuclear density map before and after matching, as can be seen in Figure 3, shows the validity of the PSM matching. Furthermore, we control for industry-province fixed effects in the matched samples, and the results are shown in column (5) of Table 8. The regression results for the matched samples are consistent with previous findings, reinforcing the robustness of the conclusions of this paper.

5. Further Analysis

The preceding analysis shows that EPT is conducive to GI. From a stakeholder perspective, we further empirically explore the heterogeneity of EPT’s impact on improving GI across various moderator variables. In addition, our exploration extends to revealing the mechanisms of EPT in contributing to economic outcomes by promoting GI.

5.1. Test of the Moderating Effect

Above, this paper has examined the mediating influence mechanism of diverse governance models on the relationship between EPT and GI. For further study, Formulas (6)–(8) are constructed by putting moderating variables and interaction terms into the benchmark Formula (1). The analysis of moderating roles on the influence of EPT on GI is conducted from the perspectives of different stakeholders, including government departments, financial markets, and suppliers. Specifically, in this section, we analyze the moderating effect of three different stakeholder channels: government subsidies, analyst coverage, and supplier concentration.

$${\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\beta }}_{0\mathrm{a}}+{\mathsf{\beta }}_{1\mathrm{a}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\beta }}_{2\mathrm{a}}{\mathrm{G}\mathrm{O}\mathrm{V}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\beta }}_{3\mathrm{a}}{\mathrm{E}\mathrm{P}\mathrm{T}\times \mathrm{G}\mathrm{O}\mathrm{V}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\beta }}_{4\mathrm{a}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(6)

$${\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\beta }}_{0\mathrm{b}}+{\mathsf{\beta }}_{1\mathrm{b}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\beta }}_{2\mathrm{b}}{\mathrm{A}\mathrm{N}\mathrm{A}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\beta }}_{3\mathrm{b}}{\mathrm{E}\mathrm{P}\mathrm{T}\times \mathrm{A}\mathrm{N}\mathrm{A}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\beta }}_{4\mathrm{b}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(7)

$${\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\beta }}_{0\mathrm{c}}+{\mathsf{\beta }}_{1\mathrm{c}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\beta }}_{2\mathrm{c}}{\mathrm{S}\mathrm{U}\mathrm{P}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\beta }}_{3\mathrm{c}}{\mathrm{E}\mathrm{P}\mathrm{T}\times \mathrm{S}\mathrm{U}\mathrm{P}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\beta }}_{4\mathrm{c}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(8)

GOV represents government subsidies as a moderator variable. ANA represents analyst coverage as another moderator variable. SUP represents supplier concentration as the last moderator variable. The set of other variables, fixed effects, and clusters are consistent with the basal regression. In the moderator effect model, we mainly focus on the coefficients of β_3a, β_3b, and β_3c, which represent the coefficient of the interaction term between EPT and moderators (i.e., EPT × GOV, EPT × ANA, and EPT × SUP), revealing the validity of the moderating effect.

5.1.1. Government Subsidies

GI behavior implies high investment and high risk, and innovation outputs may not be realized in the short term [62]. The uncertainty of innovation returns significantly affects firms’ strategic decision making [63]. To enhance GI, a combination of EPT and government subsidies is considered more effective than a single policy measure. Combined with subsidies, a carbon tax can avoid the distortion of applying a single carbon tax, and joint policies of tax and subsidies are more conducive to sustainable growth [64]. Existing research shows that government subsidies can drive firms to implement GI [65]. Firstly, government subsidies can be directly invested as resources into firms’ innovation endeavors, providing financial support and resource incentives, reducing financial costs, and dispersing R&D risks [66,67]. Secondly, government subsidies can present positive signals to market investors, so as to obtain more funding support, thereby promoting GI [68]. Due to the positive influence of subsidies on innovation funds, the incentive effect of EPT on GI can be strengthened by government subsidies. Specifically, the more pollution a firm emits, the higher the EPT it needs to pay; high-polluting firms will face higher financial pressure. Thus, the green R&D funds may be insufficient. By supporting green R&D investments through resource and signal effects, government subsidies enhance firms’ GI. Therefore, we assume that government subsidies may modulate the impact of EPT on GI.

Following [67,69], this paper selects “the logarithm of (the number of government subsidies in the annual report + 1)” and “the logarithm of (the number of government subsidies accounted in other income in the annual report + 1)” as proxy variables for government subsidies (GOV). The empirical results are shown in columns (1) and (2) of

Table 9. Test of the moderating effects.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	Government Subsidies		Analyst Coverage		Supplier Concentration
	GI	GI	GI	GI	GI	GI
EPT	−0.128 ***	−0.002	−0.037 ***	−0.038 ***	0.016 ***	0.019 ***
	(−9.51)	(−0.30)	(−7.52)	(−7.65)	(3.13)	(4.40)
GOV	−0.127 ***	−0.018 ***
	(−10.87)	(−2.63)
EPT × GOV	0.009 ***	0.002 ***
	(10.68)	(3.63)
ANA			−0.445 ***	−0.373 ***
			(−10.39)	(−10.49)
EPT × ANA			0.031 ***	0.026 ***
			(11.04)	(11.11)
SUP					0.286 *	1.384 ***
					(1.77)	(5.99)
EPT × SUP					−0.029 ***	−0.105 ***
					(−2.68)	(−6.76)
Constant	−0.779 ***	−2.873 ***	−1.796 ***	−1.774 ***	−3.173 ***	−3.234 ***
	(−3.49)	(−17.94)	(−12.95)	(−12.75)	(−19.04)	(−19.86)
Control variables	Control	Control	Control	Control	Control	Control
Province fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Industry fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations	33,220	33,220	33,220	33,220	33,220	33,220
Adjust R2	0.170	0.165	0.173	0.173	0.166	0.166

Notes: *** and * denote significance at the 1% and 10% level, respectively. T-statistics clustered at the industry-year-province level are reported in parentheses.

. The coefficients of the interaction term (EPT × GOV) are significantly positive at the 1% level, indicating that government subsidies positively regulate the relationship between EPT and GI. These findings are in line with those of [70]. Following [71], to analyze the moderating effect of government subsidies graphically, we plot the moderating effect of government subsidies under the two measures in this paper (see Figure 4). As can be seen in Figure 4, in the high government subsidies scenario, GI is significantly enhanced as firms’ EPT increases. In the context of low government subsidies, EPT has a negative or weak positive relationship with GI.

5.1.2. Analyst Coverage

Investment in GI activities has become a hot spot, and investors have been paying more attention to corporate environmental performance. Investors can directly obtain tax and investment data related to environmental protection from firms’ annual reports, but they cannot directly access the information regarding GI. Financial analysts can play the role of information intermediary and use their expertise to evaluate firms’ green investment and environmental protection projects, providing an important information channel for investors’ stock investments [72]. Existing research shows that analyst coverage can encourage firms to implement GI [73]. Firstly, financial analysts, acting as effective external governance mechanisms, alleviate information asymmetry between firms and external investors, and they broaden the ability of investors to identify information regarding firms’ GI [74]. Secondly, earnings forecast reports from financial analysts create supervision pressure on firms’ environmental governance projects. This pressure forces corporate executives to assume responsibility for environmental impact and reduce short-sighted and agency costs, and it encourages their commitment to the improvement of GI [75]. Due to the alleviation influence of information asymmetry, analyst coverage can strengthen the impact of EPT on GI. Specifically, higher EPT payments by firms attract more investor attention to environmental protection projects, and high-polluting firms face greater market supervision pressure. Consequently, enterprise executives may be inclined to abandon environmental protection projects in favor of other innovation investments in order to ease pressure. Based on the information and supervision effects, analyst coverage alleviates enterprise information asymmetry, reduces principal-agent problems, and prompts executives to prioritize the implementation of GI behaviors; therefore, we propose that analyst coverage may modulate the impact of EPT on GI.

Following [56,67], this paper selects “the logarithm of (the number of analysts following a firm + 1)” and “the logarithm of (the number of stock merchants’ research reports on a firm + 1)” as proxy variable for analyst coverage (ANA). The empirical results are shown in columns (3) and (4) of Table 9, and the coefficients of the interaction term (EPT × ANA) are significantly positive at the 1% level, indicating that the analyst coverage positively regulates the relationship between EPT and GI. These findings are consistent with the findings of [75]. The moderating effects of analyst coverage shown in Figure 5 support the robustness of our findings.

5.1.3. Supplier Concentration

The phenomenon of firms relying on suppliers is relatively common in China, where a relatively concentrated supplier base reflects supply chain stability and friendly relationships [76]. In response to green transformation, firms put forward environmental requirements for upstream suppliers, inducing the suppliers’ green motivation [77]. Examples like the claim of China’s Lujiazui Company to have purchased “toxic land” highlight the negative consequences of suppliers ignoring environmental sustainability [78]. Existing research shows that supplier concentration suppresses corporate investment in innovation [79]. Firstly, the concentration of suppliers may make the firm face the risk of operating difficulties. Once there is a supplier “bottleneck” event, supply difficulties or the high cost of switching suppliers may affect the production and operation of firms. For example, China’s Huawei company faced operational challenges due to the “United States’ ban on the supply of chips” incident. Secondly, supplier concentration causes resource dependence, constrains the bargaining power of firms with large suppliers, and hinders timely feedback on firms’ needs; thus, supplier concentration may lead to resource diversion and profit misappropriation, hindering GI. Given the impact on the risk of resource dependence, supplier concentration may impede the fulfillment of environmental responsibilities by firms, thereby depressing the positive impact of EPT on innovation. Specifically, as the attention firms attach to environmental protection increases, the green requirements for suppliers may be higher. In that case, the more concentrated suppliers are, the more likely that risk and resource effects will hinder firms’ R&D investment and impede GI. Consequently, supplier concentration is deemed as a negative regulator of the impact of EPT on GI.

Following [79,80], this paper selects the ratio of “the top five suppliers’ purchases/the total purchases” and the ratio of “the largest suppliers’ purchases/the total purchases” as proxy variables for supplier concentration (SUP). The empirical results are shown in columns (5) and (6) of Table 9. The coefficients of interaction terms (EPT × SUP) are significantly negative at the 1% level, indicating that supplier concentration negatively regulates the relationship between EPT and GI. These findings are consistent with the findings of [79]. The moderating effects of supplier concentration shown in Figure 6 support the robustness of our findings.

In summary, from the stakeholder perspective, government subsidies and analyst coverage stimulate the driving role of EPT on GI, while supplier concentration suppresses EPT’s driving effect on GI.

5.2. Test of Economic Outcomes

The diverse mechanisms of EPT on GI have been studied in the above analysis. Although many scholars have extensively examined the economic consequences of EPT and GI [59], there are few studies on whether EPT can affect economic consequences through GI; therefore, this section mainly explores the economic outcomes of EPT and GI on TFP and economic performance.

This article uses mediating effect models to reveal the influence of EPT on economic outcomes via the intermediary role of GI. Formulas (9) to (14) test mediating effects through a step-by-step approach.

$${\mathrm{T}\mathrm{F}\mathrm{P}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\beta }}_{0\mathrm{c}}+{\mathsf{\beta }}_{1\mathrm{c}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\beta }}_{2\mathrm{c}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(9)

$${\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\alpha }}_{0\mathrm{c}}+{\mathsf{\alpha }}_{1\mathrm{c}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\alpha }}_{2\mathrm{c}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(10)

$${\mathrm{T}\mathrm{F}\mathrm{P}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\gamma }}_{0\mathrm{c}}+{\mathsf{\gamma }}_{1\mathrm{c}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\gamma }}_{2\mathrm{c}}{\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\gamma }}_{3\mathrm{c}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(11)

$${\mathrm{E}\mathrm{P}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\beta }}_{0\mathrm{d}}+{\mathsf{\beta }}_{1\mathrm{d}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\beta }}_{2\mathrm{d}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(12)

$${\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\alpha }}_{0\mathrm{d}}+{\mathsf{\alpha }}_{1\mathrm{d}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\alpha }}_{2\mathrm{d}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(13)

$${\mathrm{E}\mathrm{P}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}={\mathsf{\gamma }}_{0\mathrm{d}}+{\mathsf{\gamma }}_{1\mathrm{d}}{\mathrm{E}\mathrm{P}\mathrm{T}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\gamma }}_{2\mathrm{d}}{\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{s},\mathrm{t},\mathrm{p}}+{\mathsf{\gamma }}_{3\mathrm{d}}{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{s},\mathrm{t}\mathrm{p}}+{\mathrm{I}\mathrm{n}\mathrm{d}}_{\mathrm{s}}+{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}}_{\mathrm{t}}+{\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{v}}_{\mathrm{p}}+{\mathsf{\epsilon }}_1$$

(14)

TFP represents TFP; EP represents economic performance. The set of other variables, fixed effects, and clusters are consistent with the baseline regression. In the mediator effect models, we mainly focus on the coefficients of β_1c, β_1d, α_1c, α_1d, γ_1c, γ_1d, γ_2c, γ_2d, α_1cγ_2c, and α_1dγ_2d. β_1c is the total effect of EPT on TFP; β_1d is the total effect of EPT on economic performance; α_1c and α_1d are the effects of EPT on GI; γ_1c is the direct effect of EPT on TFP; γ_1d is the direct effect of EPT on economic performance; γ_2c is the effect of GI on TFP after EPT control; and γ_2d is the effect of GI on economic performance after EPT control. α_1cγ_2c and α_1dγ_2d indicate the indirect effect of EPT transmitted through GI.

5.2.1. TFP

Pollution charges compel firms to no longer discharge pollutants at will, and EPT increases the tax burden and environmental management costs of firms. According to the Porter hypothesis, appropriate EPT can realize the “innovation compensation effect” and urge firms to carry out technological innovation to reduce production costs and improve production efficiency [81]. The following question discussed is whether EPT can further facilitate the compensation effect of GI, thereby improving firms’ TFP. Following [59], we choose the “TFP calculated using the generalized GMM method” (TFP_GMM) as the proxy variable for the firm’s TFP. Panel A of

Table 10. Test of economic outcomes.

Panel A: TFP
Variable	(1)	(2)	(3)	(4)	(5)	(6)
	TFP_GMM	GI	TFP_GMM	TFP_LP	GI	TFP_LP
EPT	0.126 ***	0.012 ***	0.126 ***	0.302 ***	0.021 ***	0.302 ***
	(10.35)	(2.86)	(10.33)	(18.05)	(5.08)	(18.04)
GI			0.023 *			0.039 *
			(1.74)			(1.94)
Constant	−5.833 ***	−3.208 ***	−5.758 ***	−8.766 ***	−2.797 ***	−8.657 ***
	(−18.33)	(−20.05)	(−17.65)	(−19.52)	(−17.77)	(−18.94)
Control variables	Control	Control	Control	Control	Control	Control
Province fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Industry fixed effects	Yes	Yes	Yes	No	No	No
Year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations	33,220	33,220	33,220	33,220	33,220	33,220
Adjust R2	0.274	0.163	0.274	0.275	0.0807	0.275
Panel B: Economic Performance
Variable	(7)	(8)	(9)	(10)	(11)	(12)
	ROA	GI	ROA	NP	GI	NP
EPT	0.014 ***	0.015 ***	0.014 ***	0.091 ***	0.016 ***	0.938 ***
	(30.99)	(3.73)	(31.05)	(2.65)	(4.17)	(21.11)
GI			0.002 ***			0.194 ***
			(4.96)			(4.13)
Constant	−0.228 ***	−3.265 ***	−0.221 ***	−4.031 ***	−3.259 ***	−15.461 ***
	(−25.60)	(−20.50)	(−24.77)	(−6.49)	(−21.05)	(−17.89)
Control variables	Control	Control	Control	Control	Control	Control
Province fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Industry fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	No	No	No
Observations	33,220	33,220	33,220	33,220	33,220	33,220
Adjust R2	0.319	0.163	0.320	0.553	0.160	0.220

Notes: *** and * denote significance at the 1% and 10% level, respectively. T-statistics clustered at the industry-year-province level are reported in parentheses.

empirically tests the influence mechanism of “EPT—GI—TFP”. We use the step-by-step mediation effect model for regression, and the results are shown in columns (1), (2), and (3). The coefficients of EPT and GI are significant, indicating that GI has a partial mediating role in the effects of EPT on TFP. This result is similar to that of [81]. The increase in GI leads to a significant increase in TFP. Specifically, appropriate EPT encourages firms to carry out green technology innovation, stimulates their demands for green technologies and green processes, and incentivizes them to improve efficiency; therefore, EPT indirectly promotes the progress of TFP through GI. Following [82], this paper also selects the “TFP calculated using LP method” (TFP_LP) for the robustness test. The regression results shown in columns (4), (5), and (6) substantiate the robustness of the findings.

5.2.2. Economic Performance

The incentivizing effect of EPT on GI is beneficial for firms to introduce talent, establish a positive corporate image among different stakeholders, improve the competitive advantage of environmental protection products, and yield economic benefits [18]. The final question remaining is whether EPT can further facilitate the compensation effect of GI to enhance firms’ economic performance. Following [18], we choose the “return of total assets” (ROA) as proxy variables for firms’ economic performance. Panel B of Table 10 empirically tests the influence mechanism of “EPT—GI—economic performance”. We use the three-step mediation effect model for regression, and the results are shown in columns (7), (8) and (9). The coefficients of EPT and GI are significant, indicating that GI has a partial mediating role in the effects of EPT on economic performance. This result is similar to that of [83]. The increase in GI leads to a significant increase in economic performance. Specifically, appropriate EPT pressure makes firms pay more attention to pollution reduction and emission reduction; to follow, the progress of GI improves the green reputation and green competitiveness of firms, thereby driving firms’ economic performance. Therefore, EPT indirectly improves a firm’s economic performance through GI. Following [36], this paper also selects “the logarithm of (net profit + 1)” (NP) for the robustness test. The regression results shown in columns (10), (11), and (12) substantiate the robustness of the findings.

The above results show that the driving effect of EPT on GI found in this paper can further improve the production efficiency and economic performance of firms.

6. Conclusions and Implications

Environmental tax collection can promote GI, but there are few empirical studies on this topic. In addition, how EPT drives GI, and whether EPT has a greater impact on GI by stakeholders, also have to be confirmed. In this research, we try to fill these research gaps.

This paper aims to evaluate the effectiveness of EPT on the GI of listed firms in China. Considering this basic goal, we use data from Shanghai and Shenzhen A-share listed firms from 2010 to 2022 to test the direct effects, mediating transmission mechanisms, moderating effects, and economic consequences of EPT on GI. The results are shown as follows: Firstly, EPT plays a significant role in promoting GI in China’s listed firms. Our results are in line with those of Zheng et al. (2023), revealing that EPT can positively promote GI [33]. After a series of robustness tests, our empirical results remain valid. Secondly, we find two indirect mechanisms affecting the relationship between EPT and GI, namely “EPT—digitalization—GI” and “EPT—ESG performance—GI”. In short, digitalization and ESG performance have a partial intermediary influence on the effects of EPT on GI. Our findings are in line with those of Jia et al. (2022) and Li et al. (2022), confirming the facilitating effects of digitalization and ESG on GI [51,84]. Thirdly, we expand our analysis from the perspective of stakeholders and find that the relationship between EPT and GI can be enhanced under the influence of government subsidies and analysts. Conversely, the relationship between EPT and GI will be suppressed under the joint influence of EPT and suppliers’ concentration. Finally, we analyze the impact of EPT on different economic consequences and find that EPT can improve enterprise TFP and economic performance through GI.

Based on the conclusions of this paper, and considering the competition situation faced by Chinese firms, this paper puts forward the following suggestions: Firstly, government entities and tax authorities should persist in refining the EPT collection system, optimizing the environmental tax system, strengthening supervision and inspection of the implementation of environmental regulations by local governments, and urging the fulfillment of environmental duties in accordance with local conditions [85]. This approach will encourage firms to accelerate their green transformation and improve their GI performance. Additionally, the government can continue to promote the digitalization and green development of industries, provide technical support for the external environment of firms’ digital governance, and improve the concept of firms’ green development. In addition, the government can strengthen the R&D support for GI and increase the support of firms that do well in terms of EPT compliance, such as providing special subsidies for firms in green transformation, reducing financing costs, and improving firms’ GI [67]. Secondly, the China Securities Regulatory Commission and financial institutions should continue to improve the construction of the analyst system and guide analysts to give full play to the advantages of information intermediaries. In addition, financial institutions can further develop green financial products, such as green credit, green bonds, and green insurance, reduce the difficulty of GI financing, and promote the green transformation of firms [86]. Thirdly, firms should attach importance to the development of digitalization, improve digital technology, improve the feedback speed of consumer demand and products, strengthen digital production and operation, promote digital governance, and promote the improvement of enterprise innovation performance [87]. At the same time, firms should take the initiative of fulfilling ESG responsibilities, disclosing details of ESG information in a timely and proactive way, establishing a positive image of environmental friendliness, and maintaining a good social reputation, thereby enhancing their market competitive position. In addition, firms should pay attention to the possible risks of relying too much on suppliers, value the relationships with suppliers, and adjust in a timely way to supply chain information.

Compared with previous research, the main contributions of this paper are shown as follows: Firstly, this paper expands the transmission mechanism of the “EPT—GI” relationship from a theoretical perspective, and it explores the bridge role of digitalization and ESG performance with the empirical tests of micro-level data. Secondly, this paper explores the joint role of EPT and various stakeholders, which also puts forward new research ideas for the co-ordinated use of different environmental regulation methods (such as EPT and government subsidies). Finally, this paper combines enterprise behavior, stakeholder behavior, and the performance of firms to enrich the research toolbox of firms’ green development. An important limitation of our study is that we use data from listed firms for analysis while ignoring small and medium firms. Future research can continue to explore the impact of EPT on GI in small and medium firms and subdivided industries.