Dividend Payments and Persistence of Firms’ Green Innovation: Evidence from China

Abstract

Green innovation is an important driver for the sustainable development of the real economy and the realization of a green transformation. Previous studies have paid less attention to the relationship between dividend payments and the persistence of green innovation and neglected the mechanism of knowledge management on the relationship between the two. Using Chinese listed non-financial companies from 2007 to 2022, this paper empirically investigates the effect and mechanism of dividend payment on corporate green innovation persistence based on two perspectives: principal–agent theory and incentive theory, using Tobit and negative binomial models. It was found that there is a non-linear inverted U-shaped relationship between dividend payments and the persistence of green innovation. Dividend payments influence the persistence of green innovation through agency costs and knowledge management capabilities. Both independent director networks and knowledge spillovers have a positive impact on the relationship between the two. Based on firms’ characteristics, the inverted U-shaped relationship between dividend payment and green innovation persistence is found to be more significant for non-state-owned firms and high-tech firms. Depending on the shareholder type, the inverted U-shaped curve between dividend payments for shareholding managers and green innovation persistence is the steepest. This study enriches the relevant literature on green innovations. In addition, it provides companies with a reference for formulating appropriate dividend policies, improving knowledge management capabilities, and improving the persistence of green innovation at a firm level.

1. Introduction

In the face of increasingly serious environmental problems, promoting the green transformation of the real economy has become one of the necessary paths for sustainable economic development [1]. Companies are the main actors of ecological environment pollution. The key to realizing green transformation is to vigorously promote the development of green innovations by enterprises, increase the output of green innovations, and improve the persistence of green innovations [2]. Green innovation is a long-term and highly uncertain business activity. The sustainability and success of green innovation activities require firms to continuously improve their willingness to innovate [3]. Under the knowledge-based and principal–agent views, firms’ knowledge resources and governance capabilities are prerequisites for increasing willingness to innovate [4,5]. Dividend policy is recognized by many scholars as a more effective incentive [6] and governance tool [7,8]. So, can dividend payments increase companies’ willingness to engage in green innovation by encouraging them to improve their knowledge management capabilities and governance, thereby promoting the persistence of green innovation?

China, one of the most polluting countries in the world, has formulated a series of targeted policies to promote industrial green transformation, with strong government intervention [9,10]. Compared to developed countries, the Chinese government promotes the transition of the economy from a planned economy to a market economy through means such as enterprise ownership reform, thereby enhancing the environmental governance level of enterprises under the combined effects of administrative and market governance [11]. At the same time, unlike the United States and the United Kingdom, the equity structure of Chinese listed companies is mostly pyramid-shaped, and the actual controllers of most enterprises are the Chinese government, which leads to differences in the scale of dividend policies implemented by listed companies [12]. In addition, due to China’s long cultural heritage, corporate culture and knowledge management are also different from those of other countries [13]. Therefore, studying the impact and mechanism of dividend payments on corporate green innovation using Chinese listed companies as the research object has certain theoretical significance and practical value.

According to the statistics from the annual reports of listed companies in China, the proportion of listed companies in the Shanghai and Shenzhen Stock Exchanges that distribute dividends increased from 50.21% in 2007 to 78.28% in 2017, and then dropped to 65.9% in 2022. The dividend payout ratio fluctuated from 20.74% in 2007 to 30.96% in 2017, and then dropped to 29.95% in 2022. The proportion of listed companies with exceptionally large dividends (the “Shenzhen Stock Exchange Listed Company Self-Discipline Supervision Guidelines No. 1—Standard Operation of Main Board Listed Companies (Revised in 2023)” issued by the Shenzhen Stock Exchange pointed out that special attention and enhanced disclosure are required when the cash dividend amount reaches or exceeds 100% of the current net profit attributable to the listed company) increased from 1.31% in 2007 to 3.33% in 2022. This indicates that the phenomena exist of non-dividend distribution, low dividend distribution, and excessive dividend distribution among listed companies in China, with the degree of excessive dividend distribution deepening over time. The existing research has shown that excessive dividend payments are a wealth transfer behavior of major shareholders and can have adverse effects on corporate operations [14]. Therefore, is there a difference in the impact of dividend payments of different scales on corporate green innovation persistence?

Based on this, this paper uses the data of non-financial listed companies in China’s A-share market from 2007 to 2022 to study the impact and mechanism of dividend payment on corporate green innovation persistence from the perspectives of agency theory and incentive theory. The marginal contributions of this paper are mainly as follows: First, it complements the relevant research on corporate green innovation. Starting from the governance and incentive ability of dividend policy, it is emphasized that dividend payment and green innovation persistence have an inverted U-shaped relationship. This confirms that a moderate dividend payment can effectively promote green innovation output and the persistence of green innovation. Second, it opens up the perspective of mechanism research. There is relatively little literature on the mechanism of dividend payment on green innovation persistence. This paper clarifies the transmission mechanism between dividend payment and green innovation in terms of both governance capacity and knowledge management capabilities. Most of the existing literature focuses only on the governance role of dividend payment and ignores its incentive effect. Based on incentive theory and knowledge management theory, this paper comprehensively examines the mechanism through which dividend payments can improve knowledge management capabilities (including intellectual capital, structural stability of intellectual capital, breadth of knowledge, environmental awareness, and willingness to innovate) by providing incentives to managers and research and development employees who hold shares and then influence the persistence of green innovations. This provides empirical evidence for a deeper understanding of the governance and incentive functions of the dividend policy on enterprises and improving corporate green innovation capabilities. Third, it shows the moderating role of independent director networks and knowledge spillovers as internal and external corporate environmental factors on the relationship between dividend payments and green innovation persistence. This provides companies with the opportunity to reduce governance costs and improve knowledge management capabilities to drive enterprise green transformation. Fourth, the different effects of dividend payments on the persistence of green innovations are analyzed in detail for companies with different types of shareholders, different ownership structures, and different industries. This represents a reference for companies with different characteristics to optimize dividend payments and promote green innovation according to local conditions.

The structure of this article is as follows: Section 2 provides the literature review. Section 3 covers theoretical analysis, which leads to the research hypothesis. Section 4 sets out the research design. Section 5 presents the empirical results and analysis. The last section contains the conclusion and policy recommendations.

2. Literature Review

2.1. Economic Consequences of Dividend Payments

The existing literature on the economic consequences of dividend payments can be broadly divided into two categories. One part of the literature argues that dividend payments are a powerful tool for corporate governance and incentives [15], and the other part of the literature argues that dividend payments can also be a self-interested tool for transferring wealth to major shareholders and management, i.e., dividend tunneling behavior [16].

Regarding the governance capabilities of dividend payments, the existing literature has concluded that firms with weaker corporate governance are more likely to reduce agency costs [17] and alleviate financing constraints [18] by paying dividends. Dividend payments can mitigate agency conflicts between managers and shareholders and between controlling and minority shareholders [19] and prevent excessive managerial investment [15,20] by reducing free cash flow within the firm. Companies can also reduce agency costs between the company and future investors by accelerating adjustments to dividend payments, thereby protecting the interests of minority shareholders and reducing financing constraints [21]. However, some studies have also found that high dividend payments can lead to a decline in companies’ internal governance capacity, causing managers to collude with large shareholders to transfer wealth within the company and increase agency costs [22]. Bond investors also fear that dividend payments will impact companies’ ability to pay interest and principal on bonds, increasing bond spreads and tightening financing constraints [23].

Regarding the incentive capabilities of dividend payments, the existing literature concludes that dividend payments, as a hidden compensation for executives, motivate managers to better maintain the company’s operations and optimize the company’s knowledge management capabilities [24,25]. Salimi and Della Torre found that dividend payments, as a collective form of compensation, can act more as an incentive when companies have low intellectual capital, which in turn promotes company innovation [26]. However, some studies have found that dividend tunneling behavior can cause large shareholders and managers to become more focused on short-term self-interests [22] and reduce their willingness to innovate, which in turn reduces corporate innovation [27]. Managers who receive excessive dividends are also more likely to transfer corporate risks to employees for their own benefit, and the incentive effect of dividend payments on managers decreases [28]. Employee-friendly companies increase future investments in the company’s employees by reducing the amount of dividend payouts, thereby reducing the incentive effect of dividend payments on employees [29].

2.2. Influencing Factors of Green Innovation

The scale of innovation capital investments and companies’ willingness to innovate are the key factors for the success of green innovation projects and an important driver for the green transformation and modernization of companies [30,31]. From the perspective of corporate governance capabilities, principal–agent problems such as executives’ on-the-job consumption and shareholders’ hollowing-out behaviors inhibit firms’ green innovation [32,33]. In terms of firms’ knowledge management capabilities, knowledge management factors such as intellectual capital stock and structural stability, green cognitive ability, and knowledge breadth also have a significant impact on firms’ green innovation [6,34,35]. In addition, previous studies have found that independent director networks, as one of the tools of internal corporate governance, can effectively perform advisory and supervisory functions, which is beneficial for companies by reducing agency costs, enhancing knowledge exchange and sharing, and promoting green innovation [36]. Knowledge spillover, as one of the ways of knowledge application and dissemination in enterprises, can improve the knowledge management capability of enterprises, enhance the willingness of R&D personnel to innovate, and promote green innovation persistence [37].

Across the literature, there is relatively extensive research in the existing studies on both the economic consequences of dividend payments and the influencing factors of green innovations, but there are still some deficits. First, existing studies have not reached a consistent conclusion about the economic consequences of dividend payments, and fewer studies have examined the impact of dividend payments on the persistence of green innovations. How dividend payments affect the persistence of green innovations is currently unknown. Second, the existing literature ignores whether independent director networks and knowledge spillovers can play a moderating role in the persistence of green innovations under dividend governance and incentive instruments. Therefore, it is necessary to include the network of independent directors and knowledge spillover in the analytical framework of dividend payments for green innovation persistence. Finally, the existing studies have paid less attention to the impact of different types of dividend payments to shareholders on the persistence of green innovations. Due to the shareholder structure, the dividends of controlling shareholders, managers, and minority shareholders differ in amount, so it is necessary to examine whether the effects of dividend payments from different types of shareholders differ on the persistence of green innovations. Therefore, from the perspective of governance capacity and knowledge management capacity, this paper clarifies the impact and mechanisms of dividend payments on green innovation persistence and discovers the “black box” of dividend payments on green innovation persistence. In this paper, the moderating effects of independent director networks and knowledge spillovers on the relationship between the two are also analyzed, as well as investigation of the heterogeneity of dividend payments on the persistence of green innovations under different types and firm characteristics. This paper enriches the literature on dividend payments and green innovation. This can help companies recognize that optimal dividend payments can serve as an effective governance tool and incentive to promote the development of green innovation. Additionally, it provides data support to encourage companies to pay moderate dividends and improve corporate governance and knowledge management capabilities.

3. Theoretical Analysis and Research Hypothesis

Enterprise green innovation has the characteristics of high risk and long cycles, which requires enterprises to provide long-term and continuous financial inputs to green innovation projects and bear the risk of green innovation failure. It has been shown that dividend payments have an impact on corporate management incentives, internal governance, and other aspects [6,38]. The relationship between dividend payments and green innovation needs to be analyzed theoretically and research hypotheses need to be drawn.

3.1. The Impact of Dividend Payments on Green Innovation Persistence

Green innovation in enterprises has the characteristics of high risk and long cycles, requiring enterprises to provide long-term and continuous funding for green innovation projects and bear the risk of green innovation failures.

Firstly, based on the agency theory, moderate dividend payments can restrain managers from expropriating internal funds, reducing the first type of agency costs [15]. Moderate dividend payments can also enhance corporate governance by constraining controlling shareholders from obtaining private benefits through internal transactions [19,38], thus curbing internal expropriation [39], safeguarding the interests of minority shareholders, and reducing the second type of agency costs. Moderate dividend payments can leverage their governance capabilities to encourage companies to allocate more funds to green innovation projects, reducing the risk of interruption or termination of green innovation projects due to funding constraints, reducing agency costs, and enhancing the persistence of green innovation.

Second, according to incentive theory, modest dividend payments as incentive bonuses can offset the costs of innovation risk for equity-involved researchers [40], attract more highly educated individuals, increase the stock of intellectual capital [26], and improve the structural stability of intellectual capital [41]. Modest dividend payments can also encourage managers to strengthen the cultivation of a green culture and increase awareness of environmental protection [6], which in turn increases the willingness of managers and R&D teams to innovate in a green way [42]. Modest dividend incentives can also improve the cohesion of R&D teams to integrate individual technical knowledge resources [25], improve the translation of technical knowledge into green patents, and expand the breadth of green knowledge [43]. Shahzad et al. found that knowledge management processes such as the acquisition, use, and dissemination of knowledge resources positively influence the persistence of green innovations [44]. Therefore, moderate dividend payments can improve the knowledge management capabilities of companies in general [45] and promote the persistence of green innovations in companies.

However, when dividend payments exceed a certain threshold, shareholders may engage in wealth transfer behavior through large dividend payouts, known as the dividend tunneling behavior [16]. Shareholders are likely to collude with executives to engage in excessive dividend payouts [22]. Excessive dividend payouts also reflect inadequate corporate governance, with executives focusing more on short-term personal gains, leading to an increase in the first type of agency costs [22]. At the same time, the incentive effect of excessive dividend payouts on R&D personnel and shareholders’ managers will be weakened, and managers are likely to no longer pay attention to promoting a green corporate culture, resulting in a decline in enterprises’ environmental perception and their willingness to innovate in a green way [27]. In addition, insufficient internal funds caused by dividend tunneling may cause researchers’ base compensation to decline. This can lead to an increase in employees’ negative emotions and a decrease in job satisfaction [46]. This may even lead to highly educated employees voluntarily leaving, a decline in the stock of intellectual capital [47], and a lack of structural stability of intellectual capital. This can further increase the risk of innovation failure and the costs of innovation [48] and hinder the persistence of green innovations. Therefore, excessive dividend payments hinder the persistence of green innovations by increasing agency costs and weakening knowledge management capabilities.

In conclusion, dividend payments may have a non-linear impact on corporate green innovation persistence. When dividend payments are below a critical threshold, the increasing dividend payments can promote the persistence of corporate green innovation through the dual channels of decreasing agency costs, and increasing the knowledge management capabilities such as intellectual capital, knowledge breadth, and green cognition. However, when dividend payments exceed the critical threshold, excessive dividend payments can inhibit the persistence of green innovation through channels such as increasing agency costs, and weakening knowledge management capabilities. Based on the above analysis, the following hypothesis is proposed in this paper:

Hypothesis 1a: 

There is an inverted U-shaped relationship between dividend payments and green innovation persistence.

Hypothesis 1b: 

Dividend payments affect green innovation persistence through the dual mechanisms of agency costs and knowledge management capabilities.

3.2. Moderating Mechanism for Independent Director Networks

Independent directors are senior management personnel who are relatively independent within the management team and possess a certain level of knowledge reserve. They bear responsibilities for consultation and supervision [49]. The existing research has shown that factors such as the proportion of independent directors [50], educational background [36], compensation incentives [51], and social networks [2] can effectively enhance the governance ability and knowledge resource provision ability of independent directors and promote the persistence of green innovation.

The higher the centrality of independent directors in the social network, the stronger their connections with other organizations in the social network [52]. In terms of consultation, independent directors in central positions can link with multiple companies through the network, thereby integrating more technical and financing information, accelerating the acquisition and integration of knowledge resources such as information [53], and mitigating information asymmetry [54]. Companies can build a knowledge interoperability platform through the network of independent directors to expand the breadth and depth of knowledge resources such as green technology innovations [36]. This increases the concentration of green knowledge capital, reduces the uncertainty of green innovation, overcomes technological barriers, and increases the probability of success in green innovation activities [2]. In terms of supervision, independent directors located in the social network are more inclined to transmit negative information about the company to stakeholders in order to enhance their reputation, reduce agency conflicts, and fulfill their supervisory role. This can improve the transparency of information in the company, enhance corporate governance [55], and promote green innovation in the company.

Therefore, when companies make moderate dividend payments, the network of independent directors and the governance and knowledge management capabilities of dividend payments come into play at the same time. This can further defuse agency conflicts, expand the breadth of knowledge, and increase companies’ willingness to participate in green innovations, which in turn strengthens the positive moderating effect of moderate dividend payouts on the promotion of green innovations. When companies pay excessive dividends, the governance and incentive abilities of dividends disappear due to the dividend tunnel effect. However, the independent director network can serve as a supplementary governance tool to mitigate the negative effects such as agency conflicts and structural instability of intellectual capital that excessive dividends may bring, thereby reducing the inhibitory effect of excessive dividends on green innovation and playing a positive regulatory role. Based on the analysis of the regulatory effects of independent director networks mentioned above, the following hypothesis is proposed in this paper:

Hypothesis 2: 

The centrality of the independent director network positively moderates the relationship between dividend payment and the persistence of green innovation.

3.3. Moderating Mechanism for Knowledge Spillover

To improve the sustainability of green innovation, companies not only need to integrate and utilize internal knowledge resources, but also rely on knowledge spillover to create a good external knowledge environment [44]. Knowledge spillover can improve communication between R&D personnel and the external knowledge environment, enriching the company’s green knowledge base and thus promoting the persistence of green innovations [37]. The stronger a company’s knowledge spillover ability, the stronger the company’s ability to disseminate and communicate in the field of technological knowledge and the stronger the company’s knowledge management ability. Companies can acquire heterogeneous green knowledge from competitors’ knowledge spillovers to overcome their own knowledge management deficiencies and further improve the persistence of green innovations.

In a favorable external knowledge environment, companies can improve information and knowledge exchange and promote the dissemination of green knowledge to other companies. These companies citing external knowledge will in turn generate new knowledge spillovers, which in turn form a cycle of knowledge spillovers between companies and promote the green transformation of the entire real economy [56]. In addition, other scholars have found that companies’ knowledge spillover capabilities can increase knowledge accumulation and managers’ trust in companies, which in turn improves companies’ knowledge management capabilities and further promotes companies’ innovation [57].

Therefore, the incentive effect of moderate dividend payouts on knowledge management capability is stronger under the influence of knowledge spillover. Increased knowledge spillover can further expand the breadth of knowledge and increase green awareness and innovation among managers. That is, knowledge spillover can strengthen the promoting effect of moderate dividend payouts on the persistence of green innovations and play a positive moderating role. When companies pay excessive dividends, their knowledge management capabilities weaken. The stronger the knowledge spillover, the more it can compensate for the loss of knowledge management capability caused by excessive dividend payouts and mitigate the inhibiting effect of excessive dividend payouts on the persistence of green innovation, which also plays a positive moderating role. Based on the above-mentioned analysis of the regulatory impact of knowledge spillover, the following hypothesis is proposed in this paper:

Hypothesis 3: 

Knowledge spillover positively moderates the relationship between dividend payments and the persistence of green innovation.

3.4. Heterogeneity Effects Based on Ownership and Industry Classification

Chinese listed companies can be divided into state-owned enterprises, with the central or local government as the actual controlling shareholder, and non-state-owned enterprises, with individuals or collectives as the actual controlling shareholders or with the state holding less than 50% of the shares. There are significant differences between state-owned enterprises and non-state-owned enterprises in terms of government support, financing constraints, corporate governance, and green innovation strategies. State-owned enterprises have closer relationships with the government and deeper political connections. State-owned enterprises with government endorsement are more likely to obtain external financing [31]. There is a greater need for non-state-owned enterprises to improve internal governance and reduce agency costs through rational dividend payouts. In terms of green innovation strategies, state-owned enterprises often bear stronger political responsibilities. As part of the policy task of accelerating the green transformation of companies, state-owned enterprises have stronger green knowledge management and are less affected by dividend payment incentives. Non-state-owned enterprises have less government intervention, and dividend payments have a greater impact on agency costs and knowledge management capabilities, which in turn have a more significant effect on green innovation persistence. Therefore, the following hypothesis is proposed:

Hypothesis 4a: 

Compared with state-owned enterprises, the impact of dividend payments by non-state-owned enterprises on green innovation persistence is more significant.

According to the list of high-tech industries published by the National Bureau of Statistics of China (the high-tech industry includes pharmaceutical manufacturing, railway, and other transportation equipment manufacturing, computer and electronic equipment manufacturing, instrument manufacturing, telecommunications and transmission services, internet and related services, software and information technology services, professional and technical services, and ecological environmental protection and environmental governance), listed companies can be divided into high-tech enterprises and non-high-tech enterprises. Zhang et al. [58] found that high-tech enterprises are more willing to implement various incentive plans for management and core technical staff. Compared to non-high-tech enterprises, high-tech enterprises, as key enterprises in the green transformation of the real economy, have a greater demand for innovation funds, a deeper reliance on technical staff, and a stronger demand for green innovation persistence. High-tech enterprises need to reduce agency costs through appropriate dividend payout, utilize incentive effects, and increase the level of environmental protection cognition, green knowledge breadth, and innovation consciousness of management and core technology employees. However, when high-tech enterprises pay excess dividends, the impact of the dividend tunnel effect on highly educated employees is greater, the loss of intellectual capital is more serious, and the adverse impact on green innovation is deeper. Therefore, the following hypothesis is proposed:

Hypothesis 4b: 

Compared with non-high-tech enterprises, the impact of dividend payments on green innovation persistence is more significant for high-tech enterprises.

3.5. Heterogeneity Effects Based on Shareholder Type

Based on the different strategic positions of shareholders in the shareholder structure and the business decisions of companies, the shareholders can be divided into controlling shareholders, shareholding managers, and minority shareholders. Based on the principal–agent theory, controlling shareholders are usually not directly involved in the company’s business decisions, but delegate management to managers on their behalf [19]. Therefore, when controlling shareholders receive dividends, they can influence the second type of agency costs [59] and slightly less in relation to knowledge management. For minority shareholders, a moderate dividend can provide an incentive to actively participate in company management. However, because minority shareholders are not directly involved in the day-to-day operations of the company, their interests can easily be violated by controlling shareholders and executives. Therefore, the dividend payment of minority shareholders has a smaller impact on agency costs and knowledge management, which in turn has a weaker impact on the persistence of green innovations.

Managers are the main decision makers in the company’s day-to-day operations and knowledge management. Compared to controlling and minority shareholders, managers who receive modest dividend payouts can curb the on-the-job consumption and directly reduce Type I agency costs. Moderate dividend payouts can also stimulate managers to increase the stock of intellectual capital and the breadth of green knowledge in the company, and increase the levels of green cognition and innovation awareness of managers [6]. Therefore, modest dividends to managers have the greatest impact in terms of reducing agency costs and improving knowledge management, thereby making them the most sensitive to the impact on the persistence of green innovations. However, when shareholding managers receive excessive dividend payments, the level of collusion between managers and controlling shareholders increases [22]. Excessive dividend payments to managers who own shares have a stronger negative impact on agency costs and knowledge management, which thereby has a stronger inhibiting effect on the persistence of green innovation. Therefore, the following hypothesis is proposed:

Hypothesis 5: 

Dividend payments by shareholding managers have the greatest influence on the persistence of green innovations compared to controlling shareholders and minority shareholders.

4. Materials and Methods

4.1. Data and Sample

Since the new accounting standards in China came into effect on 1 January 2007, in order to ensure consistency in the measurement criteria of key variables, this study used Chinese A-share listed companies from 2007 to 2022 as research samples. The green patent data were sourced from the China National Research Data Sharing Platform (CNRDS), while other data were obtained from the China Stock Market and Accounting Research (CSMAR) database. The following sample selection criteria were based on previous research practices and the needs of this study: (1) excluding samples from the financial industry; (2) removing samples from ST, PT, and delisted companies; (3) excluding samples with missing data for key variables; and (4) retaining listed companies with data for at least 3 years. In the end, a total of 1928 listed companies with 22,485 observation samples were obtained, and a 1% trim was applied to continuous data.

4.2. Measures

4.2.1. Green Innovation Persistence

Tawiah et al. [36] mentioned that from the perspective of mitigating environmental pollution, green innovation is the green optimization and improvement of the entire supply chain including the handling of raw materials, manufacturing, and the transportation of intermediate and final products. This article not only focuses on green product innovation by companies, but also considers the importance of the persistence of green innovation for corporate green transformation. Therefore, this article measures corporate green innovation persistence from the following two aspects: output of green innovation (OGI) and persistence of green innovation (PGI). Green innovation output is measured by the total number of green invention patents and green utility model patent applications. Referring to the study of Xu and Xiao [60] and Jing and Liu [61] established a method for the measurement green innovation persistence. The formula (1) is used to calculate the persistence of green innovation, as follows:

$${PGI}_t={\displaystyle \frac{{OGI}_t+{OGI}_{t-1}}{{OGI}_{t-1}+{OGI}_{t-2}}}\times \left({OGI}_t+{OGI}_{t-1}\right)$$

(1)

4.2.2. Dividend Payment

Dividend payout (DIV) is measured using the dividend payout ratio. The formula is calculated as the ratio of dividend payout per share before tax to earnings per share. To further discuss the impact of dividend payment on green innovation by different shareholder types, dividend payment is categorized into dividend payout to controlling shareholders (DIV_Controller), dividend payout to shareholding managers (DIV_Manager), and dividend payout to minority shareholders (DIV_Small) based on the structure of shareholders. Specifically, it refers to the dividends that each type of shareholder receives from the company, and is calculated as the product of the shareholding ratio of each type of shareholder and the dividend payout ratio.

4.2.3. Control Variables

This study selects control variables from the following four aspects: enterprise characteristics, enterprise operational conditions, corporate governance, and government support. The variables are age (enterprise age), size (enterprise size), SOE (state-owned enterprise), Lev (financial leverage), growth (profitability), fix (fixed asset ratio), dual (dual roles), exe (executive shareholding), and gov (government subsidies).

Table 1. Variable definition.

Variable Code	Variable Name	Variable Definition
PGI	Persistence of Green Innovation	The formula (1)
OGI	Output of Green Innovation	Green invention patent applications + green utility model patent applications
DIV	Dividend payment	Dividend per share before tax/earnings per share
DIV_Controller	Controlling shareholder dividend payment	Controlling shareholders’ shareholding × Total dividend payout/net profit
DIV_Manager	Shareholding managers dividend payment	Senior management shareholding × Total dividend payout/net profit
DIV_Small	Minority shareholders dividend payment	Shareholding by minority shareholders × Total dividend payout/Net profit
Age	Firm Age	The logarithm of the number of years since the establishment of the firm
Size	Firm Size	The logarithm of total assets.
SOE	Ownership	If it is a state-owned enterprise, it is 1, otherwise it is 0
Lev	Leverage	Total liabilities/total assets
Growth	Profitability	Net profit/total assets
Fix	Fixed Assets Ratio	Net fixed assets/total assets
Dual	Firm Leadership Structure	A dummy variable is measured to determine whether the CEO and the chairman of the board are the same person. It equals one if they are the same person; otherwise, it equals zero.
Exe	Management shareholding	The logarithm of the total shareholding of the board of directors, the board of supervisors, and senior managers
Gov	Government subsidy	The logarithm of government subsidy

provides the specific definitions of these variables.

4.3. Model

Referring to the research of Haans [62] with the conventions of the previous literature, and in order to verify that the relationship between dividend payments and green innovation is an inverted U-shape, this paper regresses dividend payments and their squared terms on green innovation persistence. Due to the lag effect of corporate green innovation, and in order to avoid the potential issue of reverse causality on the empirical results, this study uses the green innovation output (OGI) and the persistence of green innovation (PGI) in period t + 1 for regression analysis. Therefore, the following model is constructed in this study:

$${PGI}_{i,t+1}\left({OGI}_{i,t+1}\right)={\alpha }_0+{\alpha }_1{DIV}_{i,t}+{\alpha }_2{DIV}_{i,t}^2+{\alpha }_3{X}_{i,t}+{\alpha }_t+{\alpha }_j+{\epsilon }_{i,t}$$

(2)

In Equation (2), PGI_i,t+1 denotes the persistence of green innovation of company i in year t + 1; OGI_i,t+1 represents the corporate green innovation output of company i in year t + 1; DIV_i,t is the dividend payment of company i in year t; X_i,t is a vector composed of multiple control variables; α_t and α_j are time fixed effects and industry fixed effects, respectively; and ε_i,t is the random error term.

The coefficients α₁ and α₂ are the focus of this paper. If the estimated α₁ is positive and α₂ is negative and both are statistically significant, this suggests an inverted U-shaped relationship between DIV and PGI (OGI). Based on Haans [62], this article also uses the U-test in the “Baseline regression” section to test and determine the sign and significance of the slopes of α₁ + 2α₂DIV_L and α₁ + 2α₂DIV_H, whether the turning points are within the confidence intervals or not, and to further determine whether or not an inverted U-curve relationship exists between DIV and PGI (OGI).

The data characteristics of green innovation output (OGI) in enterprises are discrete integer variables. Therefore, this study uses negative binomial regression to test the model (2) [63]. Subsequently, the green innovation output is logarithmically transformed to eliminate the influence of heteroscedasticity, and the ordinary least squares method is used to retest the model (2). The data distribution of the persistence of green innovation (PGI) in enterprises exhibits left-censored truncated data characteristics. This study uses a Tobit regression model to test the model (2) [64], and ordinary least squares method is used to retest the model (2). Additionally, in order to mitigate endogeneity issues, this study employs the instrumental variable approach, Heckman two-stage test, and propensity score matching test to conduct endogeneity tests based on model (2).

5. Empirical Results and Analysis

5.1. Descriptive Statistics

Table 2. Descriptive statistics of main variables.

Variable	N	Mean	Std. Dev.	Min	Median	Max
PGI	22,485	5.600	19.745	0.000	0.000	146.286
OGI	22,485	2.086	6.971	0.000	0.000	52.000
DIV	22,485	0.300	0.325	0.000	0.248	2.000
DIV_Controller	22,485	0.108	0.126	0.000	0.072	0.698
DIV_Manager	22,485	0.021	0.054	0.000	0.000	0.340
DIV_Small	22,485	0.257	0.275	0.000	0.198	1.588
Age	22,485	2.875	0.323	1.946	2.890	3.497
Size	22,485	22.288	1.262	20.135	22.090	26.302
SOE	22,485	0.367	0.482	0.000	0.000	1.000
LEV	22,485	0.421	0.191	0.060	0.418	0.839
Growth	22,485	0.172	0.337	−0.485	0.119	1.901
Fix	22,485	0.217	0.155	0.003	0.187	0.682
Dual	22,485	0.269	0.443	0.000	0.000	1.000
Exe	22,485	13.161	6.584	0.000	15.682	20.387
Gov	22,485	8.340	8.347	0.000	12.140	20.047

reports the descriptive statistics of the variables in this study. The minimum value of DIV is 0, and the maximum value is 2.000, indicating that there is a significant disparity in dividend payments among non-financial listed companies in China, with the existence of both “no dividend” and “high dividend” phenomena. The mean of DIV is 0.300, with a median of 0.248, suggesting that the average dividend payment of non-financial listed companies in China is skewed higher by the “high dividend” companies, with half of the companies being in a low dividend state. In terms of shareholder type, the mean value of DIV_Controller is 0.108, the mean value of DIV_Manager is 0.021, and the mean value of DIV_Small is 0.257, indicating that moderate dividend payouts are more conducive to rewarding minority shareholders. The maximum value of PGI is 146.286, and the mean is 5.600. The maximum value of OGI is 52.000, the median is 0, and the mean is 2.086. This shows that there is a big gap of green innovation in listed companies in China. A total of 50% of the companies do not carry out green innovation activities.

5.2. Baseline Regression

Table 3. The impact of dividend payments on corporate green innovation.

	OLS (1)		OLS (2)		Negative Binomial (3)		Tobit (4)
	OGIt+1		PGIt+1		OGIt+1		PGIt+1
DIV	0.241 ***		2.512 ***		0.658 ***		13.254 ***
	(6.99)		(3.23)		(5.16)		(5.59)
DIV2	−0.118 ***		−1.458 ***		−0.361 ***		−7.454 ***
	(−5.42)		(−3.06)		(−4.61)		(−4.97)
Age	−0.084 ***		−1.525 ***		−0.448 ***		−6.960 ***
	(−4.22)		(−3.32)		(−6.65)		(−5.57)
Size	0.214 ***		4.758 ***		0.571 ***		12.283 ***
	(30.86)		(22.51)		(30.36)		(34.83)
SOE	0.109 ***		1.426 ***		0.401 ***		5.446 ***
	(7.61)		(4.04)		(7.80)		(5.69)
LEV	0.093 ***		−0.843		0.831 ***		6.090 ***
	(2.73)		(−1.09)		(6.61)		(2.61)
Growth	−0.031 **		−1.422 ***		−0.001		−4.127 ***
	(−2.06)		(−4.43)		(−0.02)		(−3.99)
Fix	−0.057		−3.221 ***		−0.407 ***		−5.598 *
	(−1.31)		(−3.11)		(−2.72)		(−1.94)
Dual	0.040 ***		0.936 ***		0.085 **		2.227 ***
	(3.35)		(3.18)		(2.06)		(2.88)
Exe	0.003 ***		0.031		0.005		0.331 ***
	(3.39)		(1.60)		(1.51)		(4.98)
Gov	0.000		0.013		−0.000		0.005
	(0.09)		(0.84)		(−0.23)		(0.14)
Constant	−4.519 ***		−100.452 ***		−13.891 ***		−315.222 ***
	(−28.61)		(−22.27)		(−26.21)		(−30.81)
Industry/Year FE	YES		YES		YES		YES
N	22,485		22,485		22,485		22,485
Pseudo R2	0.245		0.172		0.0948		0.0775
Turning point	1.021 ***		0.861 ***		0.912 ***		0.889 ***
Slope	0.241 ***	−0.231 ***	2.512 ***	−3.322 ***	0.628 ***	−0.785 ***	13.254 ***	−16.561 ***
U-test t-value	3.97 ***		2.68 ***		3.81 ***		4.12 ***

Note: *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively; t-values in parentheses are adjusted for heteroscedasticity. Industry FE and Year FE represent fixed effects at the industry and year levels, respectively.

presents the regression results of the impact of dividend payments on corporate green innovation using Model (2). The first column shows the OLS regression results of dividend payments on green innovation output, while the third column shows the negative binomial regression results. The coefficients of DIV² are all significantly negative, while the coefficients of DIV are all significantly positive. The results of the U-test show that the effect of DIV on OGI is first positive (slope = 2.512, p-value < 0.001) and then negative (slope = −3.322, p-value < 0.001) The turning point of DIV is 0.861, which is within the 99% confidence interval [0.522, 1.769]. Haans’ inverted U-shaped three-stage test indicates an inverted U-shaped relationship between DIV and OGI [62].

The second column displays the OLS regression results of dividend payments on persistent green innovation, while the fourth column shows the Tobit regression results. Similar to the previous analysis, the coefficients of DIV² are all significantly negative, and the coefficients of DIV are all significantly positive. The results of the U-test show that the effect of DIV on PGI is first positive (slope = 13.254, p-value < 0.001) and then negative (slope = −16.561, p-value < 0.001). The turning point of DIV is 0.889, which is within the 99% confidence interval [0.706, 1.206]. Haans’ inverted U-shaped three-stage test suggests that DIV has an inverted U-shaped relationship with PGI.

The impact of shareholder dividend payments on green innovation in enterprises follows an inverted U-shaped pattern, confirming Hypothesis 1a. This suggests that as dividend payments transition from non-existent to present, they contribute to decreasing agency costs. This helps restrict the misuse of free cash flow by executives and provides financial security for green innovation projects. Modest dividend payouts also improve knowledge management capabilities in various aspects such as intellectual capital, knowledge breadth, and environmental awareness, and promotes green innovative corporate development. The optimal dividend payment ratio for promoting green innovation output is 86.13%, and the optimal dividend payment ratio for promoting persistent green innovation is 88.91%. If the dividend payment ratio of listed companies is maintained at about 86.13% or 88.91%, the governance and incentive effects of dividend payments at this time can be maximized. The optimal dividend payment can minimize agency costs, motivate the knowledge management ability of managers, and thus promote the output of green innovations and the persistence of green innovations. However, when dividend payments increase, a dividend tunnel effect may occur and costs for Type I agencies increase. The stock and structural stability of intellectual capital are decreasing, and the success rate of green innovation activities is decreasing, which affects enterprises’ green innovation output and green innovation persistence. In this context, we will further analyze the mechanism from the aspects of agency costs and knowledge management capabilities later in the mechanism analysis section.

5.3. Endogeneity Test

In the baseline regression test, we address the issue of reverse causality between dividend payments and corporate green innovation by lagging the dependent variable by one period. We also select control variables from various aspects such as firm characteristics and operating conditions to alleviate omitted variable problems. Additionally, we reduce measurement error issues by cross-verifying data from the CNRDS data platform, WIND database, and CSMAR database. However, there are still shortcomings, so this study further addresses endogeneity issues using instrumental variable methods, the Heckman two-stage test, and propensity score matching test.

5.3.1. Instrumental Variable Method

First, referring to the instrumental variables selected by Xiao et al. [65], the average dividend payment of companies in the same industry and year in the province where the company is registered, excluding the focal company (DIV_Area) and its squared term, are used as instrumental variables. This indicator theoretically reflects the level of dividend payments in the regional industry, which is related to the dividend payments of the focal company. However, deducting the average dividend payment of the region where the company is located does not directly affect the green innovation investment decision within the company, satisfying exogeneity.

Second, we adopt the instrumental variable construction method proposed by Lewbel [66], using the difference between the dividend payments of the firm and the industry-average dividend payments in the same province and year (DIV_Lewbel) along with its cubic term as instrumental variables.

Third, we employ the shift-share method to construct Bartik instrumental variables [67]. As shown in Equation (3), we use the initial state of dividend payments and the overall growth rate to simulate the estimated values for each firm over the years, which are related to the actual values but independent of the error term, effectively meeting the relevance and exogeneity conditions, as follows:

$${DI{V}_{-}Bartik}_{i,t}={DIV}_{i,t=0}\times (1+G_{province,t})$$

(3)

Among them, DIV_Bartik_i,t is the instrumental variable for dividend payments of company i in year t; DIV_i,t=0 is the initial dividend payment of company I; and G_province,t is the dividend payment growth rate of the same industry in the province where company i is registered in year t.

Table 4. Endogeneity test—instrumental variable method.

	2SLS (1)	2SLS (2)	2SLS (3)	2SLS (4)	2SLS (5)	2SLS (6)
	IV = DIV_Area/DIV_Area2		IV = DIV_Lewbel/IV_Lewbel3		IV = DIV_Bartik/DIV_Bartik2
	OGIt+1	PGIt+1	OGIt+1	PGIt+1	OGIt+1	PGIt+1
DIV	1.448 ***	30.471 ***	0.219 ***	2.627 **	0.565 ***	12.383 ***
	(4.53)	(4.29)	(4.32)	(2.21)	(4.43)	(4.27)
DIV2	−0.826 ***	−22.332 ***	−0.103 ***	−1.559 **	−0.382 ***	−8.115 ***
	(−2.90)	(−3.63)	(−3.34)	(−2.24)	(−4.29)	(−4.19)
Controls	YES	YES	YES	YES	YES	YES
Industry/Year FE	YES	YES	YES	YES	YES	YES
N	22,485	22,485	22,485	22,485	15,324	15,324
Pseudo R2	0.063	0.020	0.108	0.082	0.122	0.091
Kleibergen–Paap rk LM	36.446		5011.114		134.873
Kleibergen–Paap rk Wald F	18.777 > 7.03		10,372.38 > 7.03		88.052 > 7.03

Note: 7.03 is the critical value for the Stock–Yogo weak instrument test at the 10% level. ** and *** indicate statistical significance at the 5% and 1% levels, respectively. t-values in parentheses are adjusted for heteroscedasticity. Industry FE and Year FE represent fixed effects at the industry and year levels, respectively.

reports the regression results of IV-2SLS estimation using the three instrumental variables mentioned above. First, the validity of the three instrumental variables is tested. The Kleibergen–Paap rk LM test results are 36.446, 5011.114, and 134.873, rejecting the unidentifiable null hypothesis at the 1% significance level. The Kleibergen–Paap rk–Wald F test results are 18.777, 10372.38, and 88.052, all greater than the critical value of 7.03 at the 10% level, rejecting the weak instrumental variable hypothesis. This indicates that the three instrumental variables selected in this study are valid. From the empirical results, the coefficient of DIV is significantly positive and the coefficient of DIV² is significantly negative, with no substantial changes. This suggests that after addressing endogeneity issues through the instrumental variables approach, the inverted U-shaped relationship between dividend payments and green innovation still holds.

5.3.2. Heckman Two-Stage Test

To mitigate sample selection bias, we chose the Heckman two-stage model for testing. In the first stage, the inverse Mills ratio (IMR) is obtained through a Probit regression model. In this stage, the dividend payment dummy variable is used as the dependent variable, defined as 1 when the dividend is greater than the industry average, and 0 otherwise. Control variables are consistent with the baseline regression model. IMR is included as a control variable in Model (2) for the second stage test. The regression results are shown in columns (1) to (2) of

Table 5. Endogeneity test—Heckman two-stage model and propensity score matching test.

	Heckman Two-Stage Model		Propensity Score Matching Test
	Negative Binomial (1)	Tobit (2)	Negative Binomial (3)	Tobit (4)
	OGIt+1	PGIt+1	OGIt+1	PGIt+1
DIV	0.659 ***	13.311 ***	0.794 ***	14.161 ***
	(5.17)	(5.61)	(5.37)	(5.21)
DIV2	−0.361 ***	−7.468 ***	−0.431 ***	−7.988 ***
	(−4.62)	(−4.98)	(−5.03)	(−4.90)
IMR	5.627 ***	59.338 *
	(3.21)	(1.83)
Constant	−29.972 ***	−485.367 ***	−14.815 ***	−342.331 ***
	(−5.95)	(−5.19)	(−23.50)	(−26.60)
Controls	YES	YES	YES	YES
Industry/Year FE	YES	YES	YES	YES
N	22,449	22,449	16,934	16,934
Pseudo R2	0.0947	0.0775	0.0999	0.0827

Note: * and *** indicate statistical significance at the 10% and 1% levels, respectively; t-values in parentheses are adjusted for heteroscedasticity. Industry FE and Year FE represent fixed effects at the industry and year levels, respectively.

, where the coefficient for DIV is significantly positive and the coefficient for DIV² is significantly negative, indicating the reliability of the results.

5.3.3. Propensity Score Matching Test

This study further adopts propensity score matching to alleviate the endogeneity issue caused by sample self-selection. Firstly, the samples are divided into two groups based on dividend payments, with the median as the threshold. Samples above the median are treated as the treatment group, while samples below the median are treated as the control group. Secondly, we select nine variables including firm age (Age), firm size (Size), leverage ratio (Lev), profitability (Growth), the shareholding ratio of the largest shareholder (SH1), board size (Board), proportion of independent directors (Indep), duality (Dual), and executive shareholding ratio (Exe) as covariates for matching. The 1:1 nearest neighbor matching method with replacement is used to match samples in the treatment group with control group samples with similar characteristics. Finally, the matched samples are regressed according to Model (2). The regression results in columns (3) to (4) of Table 5 show that the coefficients of the core variables remain unchanged, indicating the results are reliable, and the impact of dividend payments on green innovation is inverted U-shaped. The sample balance test results in

Table 6. Propensity score matching balance test.

Covariate	1:1 Matching
	%bias	t-Test
Age	0.5	0.35
Size	0.9	0.68
Lev	1.7	1.25
Growth	1.6	1.22
SH1	1.4	1.01
Board	1.9	1.47
Indep	0.1	0.06
Dual	−1.4	−1.08
Exe	−1.4	−1.02

show that the absolute values of the standard deviations of all covariates are below 10%, and the absolute values of the t-test are less than the critical value of 1.96. This indicates that there are no significant differences between the two sample groups, meeting the balance test requirements and ensuring the effectiveness of the regression.

5.4. Robustness Testing

5.4.1. Changing the Measurement Method of Core Variables

This study employs proxy variables for core variables in robustness tests. Green innovation output (OGI_g) is measured by the total number of green invention patents and green utility model patents obtained. The persistence in green innovation (PGI_rd) is measured using formula (4), where RD is the natural logarithm of the research and development (R&D) expenditures of the firm. Green innovation is closely related to innovation input, and the persistence of R&D expenditures can also represent the persistence of green innovation to some extent. Dividend payments (DIV_as) are measured by the ratio of current dividend total to total assets. The regression results, as shown in columns (1) to (2) of

Table 7. Robustness test.

	Change the Dependent Variable		Delete Municipalities		Manufacturing Firms
	Negative Binomial (1)	Tobit (2)	Negative Binomial (3)	Tobit (4)	Negative Binomial (5)	Tobit (6)
	OGI_gt+1	PGI_rdt+1	OGIt+1	PGIt+1	OGIt+1	PGIt+1
DIV_as	0.222 ***	0.548 ***
	(8.05)	(4.80)
DIV_as2	−0.021 ***	−0.058 ***
	(−6.35)	(−4.27)
DIV			0.610 ***	8.895 ***	0.641 ***	10.912 ***
			(4.39)	(3.53)	(4.50)	(4.21)
DIV2			−0.353 ***	−5.458 ***	−0.386 ***	−7.188 ***
			(−4.14)	(−3.43)	(−4.40)	(−4.35)
Constant	−15.012 ***	−115.884	−14.370 ***	−314.999 ***	−15.418 ***	−334.517 ***
	(−25.70)	(−0.09)	(−24.66)	(−28.71)	(−27.99)	(−32.96)
Controls	YES	YES	YES	YES	YES	YES
Industry/Year FE	YES	YES	YES	YES	YES	YES
N	22,485	22,117	18,028	18,028	15,193	15,193
Pseudo R2	0.107	0.101	0.101	0.0778	0.0884	0.0645

Note: *** indicate statistical significance at the 1% levels; t-values in parentheses are adjusted for heteroscedasticity. Industry FE and Year FE represent fixed effects at the industry and year levels, respectively.

, maintain consistency with the previous text. This indicates that regardless of the replacement of explanatory or explanatory variables, the results of this study remain robust.

$${PGI\_rd}_t={\displaystyle \frac{{RD}_t+{RD}_{t-1}}{{RD}_{t-1}+{RD}_{t-2}}}\times \left({RD}_t+{RD}_{t-1}\right)$$

(4)

5.4.2. Adjusting the Observation Sample

First, delete the samples of enterprises registered in municipalities directly under the central government. Due to the obvious location advantages of China’s municipalities directly under the central government, there are significant differences in terms of economy, population, and government resources compared to other regions. Enterprises located in municipalities directly under the central government may have differences in daily operations, financing, and investment compared to other enterprises, which may impact the benchmark regression results. Therefore, this study deleted samples of enterprises registered in Beijing, Chongqing, Shanghai, and Tianjin, resulting in 18,028 samples for robustness testing.

Second, only retain the samples of manufacturing enterprises. The core of the national real economy green transformation is the manufacturing industry. Deleting other industries can eliminate industry factors that may not be observable, making the regression results more representative. Therefore, this study only retained samples of the manufacturing industry, resulting in 15,193 samples for robustness testing. The regression results are shown in columns (3) to (6) of Table 7, where the coefficient of DIV is significantly positive and the coefficient of DIV² is significantly negative, indicating that the results of this study are robust and further validate Hypothesis 1a.

5.5. Mechanism Analysis

The baseline regression results confirm that the relationship between dividend payments and corporate green innovation is inverted U-shaped. Reasonable dividend payments can effectively promote green innovation, while excessive dividend payments significantly reduce green innovation. Combining the theoretical analysis of dividend payments on corporate green innovation in the previous section, mechanism testing can be conducted from the perspectives of signal transmission effects, agency costs, and green innovation willingness.

When dividend payments are at a low level, they can send positive signals to the outside world, reduce financing constraints, decrease agency costs, and provide financial support for green innovation. Moderate dividend payouts can also generate incentive effects, enhance human capital and management’s environmental awareness, increase willingness for green innovation, and promote green innovation. When companies engage in “excessive dividend payouts,” shareholders may exhibit dividend tunneling behavior, negative signals leading to increased financing costs and higher agency costs. Dividend tunneling behavior can also result in loss of human capital, increased risks of failure in green innovation, decreased willingness for innovation among management, and ultimately inhibit green innovation. This study examines the impact paths and mechanisms of dividend payments on green innovation based on the above logic. Since the effects of the above mechanisms may vary in different scales of dividend payments, this paper divides the full sample into two sub-samples, the low dividend group and high dividend group, based on the mean dividend payment, and sets the regression model as follows:

$$M_{i,t+1}=\left\{\begin{array}{c}{\beta }_0+{\beta }_1{DIV}_{i,t}+{\beta }_2{X}_{i,t}+{\alpha }_t+{\alpha }_j+{\epsilon }_{i,t} ,{ DIV}_{i,t}\le DI{V}_{mean} \\ {\gamma }_0+{\gamma }_1{DIV}_{i,t}+{\gamma }_2{X}_{i,t}+{\alpha }_t+{\alpha }_j+{\epsilon }_{i,t} ,{ DIV}_{i,t}>DI{V}_{mean} \end{array}\right.$$

(5)

In order to alleviate the endogeneity issue caused by mutual causality, the instrumental variable is set as t + 1 period. M_i,t+1 represents the instrumental variable of company i in year t + 1. Other indicators are defined similarly as in Equation (2).

5.5.1. Mechanism Examination Based on Agency Costs

Reviewing the theoretical analysis of dividend payments on corporate green innovation in the preceding text, we find that dividend payments are closely related to agency costs. The first type of agency costs manifest as management extracting private benefits through on-the-job consumption, leading to agency conflicts with shareholders. The second type of agency costs manifest as controlling shareholders obtaining private benefits through fund appropriation, related-party transactions, etc., resulting in agency conflicts with minority shareholders. Referring to Luo et al. [68], the first type of agency costs is measured by on-the-job consumption (AC), with the formula as follows: on-the-job consumption = (office expenses + travel expenses + communication expenses + entertainment expenses + overseas training expenses + car expenses + meeting expenses)/operating income. Referring to Bian et al. [22], the second type of agency costs is measured by related-party transactions (RPT), with the formula as follows: related-party transactions = total amount of related-party transactions/total assets.

The examination of the mechanism of dividend payments on agency costs is shown in

Table 8. Mechanism analysis—agency costs.

	DIV_Low	DIV_High	DIV_Low	DIV_High
	OLS(1)	OLS (2)	OLS (3)	OLS (4)
	ACt+1		RPTt+1
DIV	−0.554 ***	0.084 ***	−0.230 ***	0.009
	(−8.75)	(3.77)	(−7.68)	(0.81)
Constant	4.465 ***	2.935 ***	0.850 ***	0.684 ***
	(13.85)	(10.88)	(8.79)	(5.50)
Controls	YES	YES	YES	YES
Industry/Year FE	YES	YES	YES	YES
N	12,104	8480	12,347	9001
R2	0.262	0.261	0.164	0.178

Note: *** indicate statistical significance at the 1% levels; t-values in parentheses are adjusted for heteroscedasticity. Industry FE and Year FE represent fixed effects at the industry and year levels, respectively.

. The coefficients of the low dividend group (DIV) in columns (1) and (2) of the regression results on in-service consumption are significantly negative, indicating that when dividend payments are at a lower level, increasing dividend payments can effectively constrain executives’ self-interest behavior, leading to a decrease in in-service consumption. The coefficients of the high dividend group (DIV) are significantly positive, indicating that the higher the excess dividend payments, the more severe the wealth transfer behavior of shareholders, the lower the level of corporate governance, and the more it helps to boost executives’ self-interested behavior, leading to an increase in in-service consumption.

The results of the regression analysis on dividend payments and related party transactions are presented in columns (3) and (4) of Table 8. The coefficients for the low dividend group DIV are all significantly negative, while the coefficients for the high dividend group DIV are not significant. This indicates that when dividend payments are at a low level, increasing dividend payments can effectively limit the size of related party transactions, reduce the interests of controlling shareholders in expropriating minority shareholders, and decrease agency costs. When dividend payments are excessive, they no longer have a significant impact on related party transactions within the firm. This may be because shareholders directly extract internal funds through dividend tunneling behavior, eliminating the need for indirect methods such as related party transactions.

In summary, moderate dividend payments can significantly reduce in-service consumption, related transactions, and suppress agency costs. Excessive dividend payments, on the other hand, significantly increase on-the-job consumption and the first type of agency costs rise. Combined with the existing literature confirming that agency costs adversely affect corporate green innovation [32,33], this subsection verifies that dividend payments affect green innovation persistence through agency costs, verifying Hypothesis 1b. Specifically, moderate dividend payouts reduce Type I and Type II agency costs, which in turn promote green innovation persistence. Excessive dividend payouts, on the other hand, cause dividend tunneling effects and increase agency costs, which in turn inhibit green innovation persistence.

5.5.2. Mechanism Examination Based on Knowledge Management

Based on knowledge management theory, knowledge management capability can be defined as the creation, acquisition, integration, use, and dissemination of knowledge resources by organizations [69]. Intellectual capital is an important prerequisite for companies to exploit their knowledge management capabilities, and the stability of the stock and structure of intellectual capital can, to a certain extent, determine the creation ability and use efficiency of knowledge resources [70]. At the same time, the degree of environmental awareness, the breadth of green knowledge, and the innovation awareness of enterprises represent the enterprise’s ability in green knowledge management and technical knowledge management. Therefore, this article chooses the five indicators of intellectual capital stock, stability of intellectual capital structure, environmental awareness, the breadth of green knowledge, and innovation awareness as the proxy variables for knowledge management ability to more comprehensively investigate the mechanism of knowledge management ability.

According to Yue [48], intellectual capital is measured by the ratio of the sum of master’s and doctoral staff to total staff (IC).

Table 9. Mechanism analysis—stock and structural stability of intellectual capital.

	DIV_Low	DIV_High	DIV_Low	DIV_High
	OLS (1)	OLS (2)	OLS (3)	OLS (4)
	ICt+1		SICt+1
DIV	0.007	−0.004 ***	0.118 ***	−0.008 **
	(1.29)	(−2.82)	(11.39)	(−2.35)
Constant	−0.066 ***	−0.032 *	0.832 ***	0.937 ***
	(−3.81)	(−1.94)	(26.56)	(23.75)
Controls	YES	YES	YES	YES
Industry/Year FE	YES	YES	YES	YES
N	9442	6714	13,066	9416
R2	0.261	0.279	0.038	0.040

Note: *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively; t-values in parentheses are adjusted for heteroscedasticity. Industry FE and Year FE represent fixed effects at the industry and year levels, respectively.

presents the regression results of dividend payments on intellectual capital in columns (1) and (2). The coefficient for the low dividend group (DIV) is not significant, while the coefficient for the high dividend group (DIV) is significantly negative. This suggests that moderate dividend payouts have less impact on intellectual capital. On the other hand, companies that pay excessive dividends lead to a decline in the proportion of highly qualified workers and a greater loss of intellectual capital.

Referring to Cao and Yu [34], model (6) is used to measure the structural stability of intellectual capital (SIC), as follows:

$${SIC}_{t,t+1}={\displaystyle \frac{E_t-\#(S_t/S_{t+1})}{E_t}}\times {\displaystyle \frac{E_{t+1}}{E_t{+E}_{t+1}}}+{\displaystyle \frac{E_{t+1}-\#(S_{t+1}/S_t)}{E_{t+1}}}\times {\displaystyle \frac{E_t}{E_t{+E}_{t+1}}}$$

(6)

In the above equation, E_t and E_t+1 are the total number of master’s and doctoral employees in year t and year t + 1; #(S_t/S_t+1) is the number of the company’s highly qualified employees who are in the job in year t, but leave the job in year t + 1; and #(S_t+1/S_t) is the number of highly qualified employees of the company who are new to the job in year t + 1. The larger the value of SIC, the more stable the structure of intellectual capital.

Columns (3) and (4) of Table 9 show the results of the regression analysis linking dividend payouts to the robustness of the intellectual capital framework. It is noteworthy that the DIV variable has a significantly positive correlation in the group with lower dividend payouts, while in the group with higher dividend payouts, the DIV coefficient is significantly negative. This suggests that a balanced dividend policy can strengthen the structural integrity of intellectual capital and harness the motivational potential of such a policy. Conversely, companies that overpay dividends tend to undermine the structural soundness of intellectual capital and impair their knowledge management capabilities.

According to Gadenne et al. [42], the environmental awareness (EA) of enterprises was measured using content analysis. If relevant indicators (the eight relevant indicators of executive environmental awareness include environmental concepts, environmental goals, environmental management system, environmental education and training, environmental special actions, environmental emergency response mechanisms, environmental honors or awards, and the system of “simultaneous design, construction, and operation of environmental protection facilities and main projects”) were disclosed in the annual reports or corporate social responsibility reports of listed companies, they were assigned a value of 1, otherwise 0. The environmental awareness was the sum of all indicator scores, which were then normalized. The regression results of dividend payments on executive environmental awareness are shown in columns (1) and (2) of

Table 10. Mechanism analysis—environmental awareness, green knowledge breadth and innovation awareness.

	DIV_Low	DIV_High	DIV_Low	DIV_High	DIV_Low	DIV_High
	OLS (1)	OLS (2)	OLS (3)	OLS (4)	OLS (5)	OLS (6)
	EAt+1		GKBt+1		IAt+1
DIV	0.202 ***	−0.021 ***	0.163 ***	0.001	0.001 ***	−0.000 *
	(9.01)	(−2.62)	(5.97)	(0.14)	(5.17)	(−1.76)
Constant	−1.966 ***	−2.025 ***	−1.092 ***	−1.544 ***	0.015 ***	0.012 ***
	(−24.03)	(−21.81)	(−14.32)	(−17.02)	(17.07)	(10.61)
Controls	YES	YES	YES	YES	YES	YES
Industry/Year FE	YES	YES	YES	YES	YES	YES
N	12,930	9325	12,364	8967	12,866	9314
R2	0.257	0.277	0.198	0.255	0.329	0.375

Note: * and *** indicate statistical significance at the 10% and 1% levels, respectively; t-values in parentheses are adjusted for heteroscedasticity. Industry FE and Year FE represent fixed effects at the industry and year levels, respectively.

. The coefficient of the low dividend group (DIV) is significantly positive, while the coefficient of the high dividend group (DIV) is significantly negative. This indicates that moderate dividend payments can motivate executives to strengthen the green culture of the firm and increase their environmental awareness, but excessive dividend payments lead to a decrease in environmental awareness.

Drawing on Wang’s [35] measurement of knowledge breadth, the following formula is used to measure the green knowledge breadth (GKB) of an enterprise based on the IPC (International Patent Classification) main classification number of green patent applications in China:

$${GKB}_{it}=1-\sum {\left({\displaystyle \frac{Z_{imt}}{Z_{it}}}\right)}^2$$

(7)

where Z_imt is the cumulative number of green patent applications (including green invention and green utility model patents) in IPC’s m major groups for firm i up to year t and Z_it is the cumulative number of green patent applications for firm i up to year t. The larger the GKB, the larger the green knowledge breadth of the firm.

Columns (3) and (4) of Table 10 present the results of the regression of dividend payment on breadth of green knowledge. The coefficient of the low dividend group (DIV) is significantly positive, while the coefficient of the high dividend group (DIV) is not significant. This suggests that moderate dividend payouts can increase the breadth of companies’ green knowledge and improve the ability to create and use green technological knowledge. This means that knowledge management skills are improved, which in turn promotes the persistence of green innovations. On the other hand, excessive dividend payments have less impact on the breadth of green knowledge.

Drawing on the research of Chen [71], text analysis was used to calculate the total number of keywords including “R&D” and “innovation” in the Management Discussion and Analysis (MD&A) sections of the companies’ annual reports that reflect the executives’ sense of innovation. The natural logarithm of this total number of words is then used to measure executive innovation awareness (IA). Columns (5) and (6) of Table 10 show the regression results of dividend payments on executive innovation awareness. The coefficient of DIV is significantly positive for the low dividend payout group and significantly negative for the high dividend payout group. It indicates that moderate dividend payout can motivate executives to increase their innovation awareness, but excessive dividend payout leads to a decrease in the innovation awareness of executives.

In summary, moderate dividend payment can significantly improve the structural stability of intellectual capital, increase environmental awareness, the breadth of green knowledge and innovation awareness of enterprises, and comprehensively improve the knowledge management ability of enterprises. Moderate dividend payments can play an effective role in creating incentives for shareholders and R&D personnel to promote the persistence of green innovation by improving knowledge management capabilities. An excessive dividend payment, on the other hand, reduces the stock of intellectual capital and structural stability and weakens the environmental awareness and innovative strength of companies. Overpayment of dividends leads to the disappearance of the incentive effect of dividend payments and inhibits the persistence of green innovation by reducing knowledge management capabilities. In this subsection, the role of knowledge management capability in the expectation of mechanism transfer between dividend payments and persistence of green innovations is tested and Hypothesis 1b is confirmed.

5.6. Moderation Effect Analysis

Due to the impact of dividend payments on green innovation persistence being inverted U-shaped, in order to more intuitively examine the moderating effects of independent director networks and knowledge spillover on the relationship between dividend payments and green innovation persistence, the sample is divided into the low dividend group (DIV_Low) and high dividend group (DIV_High) based on the mean of dividend payments, and the following moderation effect model (8) is constructed, where Vit is the moderating variable of company i in year t, and other indicators are defined the same as in formula (2), as follows:

$${PGI}_{i,t+1}\left({OGI}_{i,t+1}\right)=\left\{\begin{array}{c}{\phi }_0+{\phi }_1{DIV}_{i,t}+{\phi }_2{V}_{i,t}+{{\phi }_{3}DIV}_{i,t}\times V_{i,t}+{\phi }_4{X}_{i,t}+{\alpha }_t+{\alpha }_j+{\epsilon }_{i,t},{ DIV}_{i,t}\le DI{V}_{mean} \\ {\omega }_0+{\omega }_1{DIV}_{i,t}+{\omega }_2{V}_{i,t}+{{\omega }_{3}DIV}_{i,t}\times V_{i,t}+{\omega }_4{X}_{i,t}+{\alpha }_t+{\alpha }_j+{\epsilon }_{i,t},{ DIV}_{i,t}>DI{V}_{mean} \end{array}\right.$$

(8)

5.6.1. Moderating Effect of Independent Director Networks

Referring to Qiu and Yu [2], the measurement method of independent director networks is used, and the centrality of the independent director network is calculated using formula (9), as follows:

$${Degree}_i={\sum }_j{N}_{ji}/g-1$$

(9)

In this context, i represents a certain independent director, while j represents other independent directors excluding i in the current period. Nji is a network link, which equals 1 if director i and director j serve together on the board of directors of at least one company, and 0 otherwise.

The regulatory effect of independent director networks is empirically demonstrated in

Table 11. Moderating effects—independent director networks.

	DIV_Low	DIV_High	DIV_Low	DIV_High
	Negative Binomial (1)	Negative Binomial (2)	Tobit (3)	Tobit (4)
	OGIt+1		PGIt+1
DIV	0.217 **	−0.087 ***	19.821 ***	−8.317 ***
	(2.27)	(−2.67)	(3.05)	(−3.54)
Degree	−0.122 ***	−0.127 ***	−3.625 *	−7.715 ***
	(−4.58)	(−2.93)	(−1.87)	(−2.84)
DIV × Degree	0.636 ***	0.192 ***	8.691	12.923 ***
	(3.56)	(2.86)	(0.72)	(3.26)
Constant	−3.602 ***	−5.409 ***	−266.345 ***	−369.682 ***
	(−17.20)	(−21.56)	(−19.91)	(−21.43)
Controls	YES	YES	YES	YES
Industry/Year FE	YES	YES	YES	YES
N	12,721	9155	12,721	9155
Pseudo R2	0.224	0.302	0.0690	0.0977

Note: *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively; t-values in parentheses are adjusted for heteroscedasticity. Industry FE and Year FE represent fixed effects at the industry and year levels, respectively.

. The results of the moderation effect of independent director networks on dividend payments and corporate green innovation output are shown in columns (1) and (2) of Table 11, where the coefficient of DIV × Degree is significantly positive in both the low dividend and high dividend groups. This indicates that when companies pay moderate dividends, the advisory capacity of independent director networks and the governance ability of dividend payments work together. The advisory capacity of a network of independent directors improves the ability to integrate and utilize the firm’s knowledge resources, while the governance capacity further reduces the firm’s agency costs. When a company overpays its dividend, the monitoring ability of the network of independent directors comes into play to mitigate the negative impact of dividend tunneling behavior on the company. The results of the moderation effect test on the relationship between dividend payments and persistence of green innovation by the centrality of the independent director network are presented in columns (3) and (4) of Table 11. The coefficient of DIV × Degree is not significant in the low dividend group but is significantly positive in the high dividend group. This indicates that a higher centrality of the independent director network strengthens the supervisory governance capacity in terms of the sustainability of corporate green innovation when there is an excessive dividend payout. The independent director network can alleviate the inhibitory effect of excessive dividend payments on persistence of green innovation. In summary, the network of independent directors strengthens the positive effects of moderate dividend payment on green innovation and weakens the negative effect of excessive dividend payment on green innovation. That is, the independent director network positively moderates the relationship between dividend payment and green innovation persistence, thereby confirming hypothesis H2.

5.6.2. Moderating Effect of Knowledge Spillover

Following Arora [72], knowledge spillover represents the dissemination ability of a company’s green patent knowledge, i.e., the citation of a company’s green patent. If a company’s green patent is cited more frequently by other people, it indicates that the company’s ability to disseminate green knowledge is stronger. Therefore, this paper uses the logarithm of the number of times a green invention patent is cited plus one to calculate a firm’s knowledge spillover.

The empirical results of the moderating effect of knowledge spillovers are presented in

Table 12. Moderating effects—knowledge spillover.

	DIV_Low	DIV_High	DIV_Low	DIV_High
	Negative Binomial (1)	Negative Binomial (2)	Tobit (3)	Tobit (4)
	OGIt+1		PGIt+1
DIV	0.046	−0.018	−0.460	−2.797
	(0.69)	(−1.12)	(−0.20)	(−1.52)
KS	0.309 ***	0.411 ***	5.964 ***	16.671 ***
	(23.95)	(27.04)	(28.63)	(27.65)
DIV × KS	0.387 ***	0.006	6.112 ***	0.059
	(4.11)	(0.24)	(4.18)	(0.07)
Constant	−0.521 ***	−1.310 ***	−13.541 ***	−130.135 ***
	(−2.75)	(−6.95)	(−2.81)	(−8.64)
Controls	YES	YES	YES	YES
Industry/Year FE	YES	YES	YES	YES
N	10,691	10,316	10,691	10,316
Pseudo R2	0.400	0.513	0.033	0.142

Note: *** indicate statistical significance at the 1% levels; t-values in parentheses are adjusted for heteroscedasticity. Industry FE and Year FE represent fixed effects at the industry and year levels, respectively.

. Columns (1) and (2) in Table 12 are the results of the moderating effect test of green knowledge spillovers on dividend payments and firms’ green innovation output. Columns (3) and (4) in Table 12 present the results of the moderating effects of green knowledge spillovers on dividend payments and the persistence of firms’ green innovations. The coefficients of DIV × KS are all significantly positive in the low dividend payout group, and none of the coefficients are significant in the high dividend payout group.

This suggests that when a firm pays moderate dividends, the incentive effect of dividend payouts on shareholding managers and research and development personnel and the firm’s knowledge spillover capacity interact. As the knowledge spillover capacity increases, the knowledge management capacity created by moderate dividend distribution also increases, thereby promoting the persistence of green innovations. However, when firms pay too high dividends, the moderating effect of knowledge spillover capacity is not significant. The reason for this may be that the overpayment of dividends leads to a decline in knowledge management capability, and the increase in knowledge spillover capability of enterprises cannot compensate for the decline in various aspects of knowledge management capability. Therefore, the moderating effect of knowledge spillover capability on the relationship between excess dividend payout and green innovation persistence is small. In summary, the spillover effects of green knowledge positively moderated the effect of dividend payments on the persistence of green innovations, confirming Hypothesis 3.

5.7. Heterogeneity Analysis

5.7.1. Grouping Study Based on Ownership of Enterprises

Table 13. Heterogeneity empirical results—corporate ownership.

	State-Owned Enterprises	Non-State-Owned Enterprises	State-Owned Enterprises	Non-State-Owned Enterprises
	Negative Binomial (1)	Negative Binomial (2)	Tobit (3)	Tobit (4)
	OGIt+1		PGIt+1
DIV	0.496 **	0.836 ***	13.304 ***	12.849 ***
	(2.24)	(5.29)	(2.95)	(4.78)
DIV2	−0.270 **	−0.433 ***	−6.902 **	−7.335 ***
	(−2.00)	(−4.48)	(−2.44)	(−4.30)
Constant	−14.958 ***	−14.885 ***	−365.525 ***	−292.752 ***
	(−17.19)	(−19.97)	(−20.13)	(−21.57)
Controls	YES	YES	YES	YES
Industry/Year FE	YES	YES	YES	YES
N	8244	14,241	8244	14,241
Pseudo R2	0.104	0.102	0.0999	0.0724

Note: ** and *** indicate statistical significance at the 5% and 1% levels, respectively; t-values in parentheses are adjusted for heteroscedasticity. Industry FE and Year FE represent fixed effects at the industry and year levels, respectively.

presents the results of the grouped regression based on Model (2) using the nature of ownership as a variable. Columns (1) and (3) display the regression results of state-owned enterprises’ dividend distribution on green innovation output and the persistence of green innovation. The coefficient of DIV is significantly positive at the 5% level, while the coefficient of DIV² is significantly negative at the 5% level. Columns (2) and (4) show the regression results of non-state-owned enterprises’ dividend distribution on green innovation output and the persistence of green innovation. The coefficient of DIV is significantly positive at the 1% level, while the coefficient of DIV² is significantly negative at the 1% level. In terms of significance level, the coefficient of dividend distribution for non-state-owned enterprises is more significant. Moreover, the coefficient values of dividend distribution for non-state-owned enterprises are higher than those for state-owned enterprises. This shows that the dividend payment of non-state-owned firms has a more significant effect on green innovation persistence compared to state-owned firms, verifying Hypothesis 4a.

5.7.2. Grouping Study Based on Industry Technological Characteristics

Table 14. Heterogeneity empirical results—technological characteristics.

	High-Tech Enterprises	Non-High-Tech Enterprises	High-Tech Enterprises	Non-High-Tech Enterprises
	Negative Binomial (1)	Negative Binomial (2)	Tobit (3)	Tobit (4)
	OGIt+1		PGIt+1
DIV	1.411 ***	0.366 **	24.509 ***	7.827 ***
	(5.94)	(2.38)	(5.74)	(2.74)
DIV2	−0.746 ***	−0.231 **	−13.026 ***	−4.869 ***
	(−5.03)	(−2.47)	(−4.75)	(−2.72)
Constant	−13.598 ***	−14.355 ***	−281.313 ***	−334.837 ***
	(−16.20)	(−23.38)	(−17.27)	(−28.62)
Controls	YES	YES	YES	YES
Industry/Year FE	YES	YES	YES	YES
N	6252	16,233	6252	16,233
Pseudo R2	0.0818	0.0998	0.0544	0.0876

Note: ** and *** indicate statistical significance at the 5%, and 1% levels, respectively; I-values in parentheses are adjusted for heteroscedasticity. Industry FE and Year FE represent fixed effects at the industry and year levels, respectively.

presents the results of the grouped regression based on enterprise technological characteristics using Model (2). The first and third columns show the regression results of the dividend payments of high-tech enterprises on green innovation output and the persistence of green innovation. The coefficient of DIV is significantly positive at the 1% level, while the coefficient of DIV² is significantly negative at the 1% level. The second and fourth columns show the regression results of dividend distribution of non-high-tech enterprises on green innovation output and the persistence of green innovation. The coefficients of DIV and DIV² on green innovation output are significant at the 5% level, and the coefficients on the persistence of green innovation are significant at the 1% level. At a significant level, compared to non-high-tech enterprises, the impact of dividend payments of high-tech enterprises on green innovation output is more significant. However, the significance of the impact of dividend payments of both types of enterprises on the persistence of green innovation is the same. In terms of coefficient values, the coefficients of dividend distribution of high-tech enterprises are greater than those of non-high-tech enterprises. This indicates that compared to non-high-tech enterprises, the impact of dividend payments of high-tech enterprises on the persistence of green innovation is more significant, confirming Hypothesis 4b. Compared to non-high-tech enterprises, reasonable dividend payments of high-tech enterprises is more conducive to reducing agency costs, and enhancing the knowledge management capability of enterprises. The excessive dividend distribution of high-tech enterprises has a deeper impact on management and core technical staff, and a stronger inhibitory effect on green innovation. The relationship between dividend distribution and green innovation persistence of high-tech enterprises follows a steep inverted U-shaped curve.

5.7.3. Grouping Study Based on Shareholder Type

The preceding mechanism test shows that dividend payment affects green innovation through agency costs and the knowledge management capability. Moderate dividend payment can effectively inhibit executives’ on-the-job consumption, manipulation of connected transactions by controlling shareholders, and increase the knowledge management capability, which in turn promotes green innovation persistence. Excessive dividend payout, on the other hand, increases executives’ on-the-job consumption and reduces the knowledge management capability, thus inhibiting green innovation persistence. Therefore, based on shareholder type, controlling shareholders, shareholding executives, and small and medium-sized shareholders may have differential impacts on the persistence of green innovation when receiving dividend payments of different sizes.

Table 15. Heterogeneity empirical results—shareholder type.

	Negative Binomial (1)	Negative Binomial (2)	Negative Binomial (3)	Tobit (4)	Tobit (5)	Tobit (6)
	OGIt+1	OGIt+1	OGIt+1	PGIt+1	PGIt+1	PGIt+1
DIV_Controller	1.561 ***			30.724 ***
	(4.39)			(4.66)
DIV_Controller2	−2.175 ***			−45.300 ***
	(−3.38)			(−3.78)
DIV_Manager		4.265 ***			51.274 ***
		(4.53)			(2.90)
DIV_Manager2		−11.414 ***			−148.467 **
		(−3.55)			(−2.43)
DIV_Small			0.721 ***			15.290 ***
			(4.65)			(5.29)
DIV_Small2			−0.486 ***			−10.104 ***
			(−4.03)			(−4.37)
Constant	−13.884 ***	−14.290 ***	−13.819 ***	−315.466 ***	−320.958 ***	−313.343 ***
	(−26.19)	(−26.96)	(−25.98)	(−30.83)	(−31.26)	(−30.54)
Controls	YES	YES	YES	YES	YES	YES
Industry/Year FE	YES	YES	YES	YES	YES	YES
N	22,485	22,485	22,485	22,485	22,485	22,485
R2	0.0947	0.0947	0.0947	0.0774	0.0773	0.0775

Note: ** and *** indicate statistical significance at the 5% and 1% levels, respectively; t-values in parentheses are adjusted for heteroscedasticity. Industry FE and Year FE represent fixed effects at the industry and year levels, respectively.

presents the empirical results of different types of dividend payments on green innovation persistence. Columns (1) to (3) show the regression results of dividend payments by controlling shareholders, shareholding executives, and small and medium-sized shareholders on green innovation output. In terms of significance, there is no significant difference among the three types of dividend payments. However, in terms of coefficient values, dividend payout by shareholding managers (DIV_Manager) has the largest impact on green innovation output. As shown in the left panel in Figure 1, the inverted U-shaped curve between dividend payouts from shareholding managers and green innovation output is the steepest. Columns (4) to (6) present the regression results of different types of dividend payments on the persistence of green innovation. In terms of significance, compared to the coefficients on dividend payments to controlling and small and medium shareholders, which are significant at the 1% level, the significance of dividend payouts by shareholding managers is weaker, being significant only at the 5% level. However, in terms of coefficient values, dividend payout by shareholding managers has the largest impact on the persistence of green innovation. As shown in the right panel in Figure 1, the inverted U-shaped curve between shareholding executive dividends and the persistence of green innovation is the steepest. In summary, based on shareholder type, dividend payouts by shareholding managers have the most sensitive impact on green innovation, with the steepest inverted U-shaped curve, confirming Hypothesis 5. The reason may lie in the fact that managers are the main decision makers in the business activities and knowledge management of the firm compared to controlling and small and medium shareholders. Dividend payments have a deeper impact on executives’ agency costs and knowledge management capabilities, which in turn has a stronger effect on the promotion of green innovation persistence.

6. Conclusions

Based on the data of non-financial listed A-share companies from 2007 to 2022, this paper examines the impact of dividend payment on the persistence of corporate green innovation and its role as a mechanism from the agency costs and knowledge management dimensions. It turns out that the effects of dividend payment on both output and persistence of green innovations has an inverted U-shape. Moderate dividend payment can promote corporate green innovation persistence by minimizing agency costs, and raising the knowledge management capability. Excessive dividend payments can affect the persistence of companies’ green innovations by increasing agency costs and decreasing knowledge management capabilities. The moderating effect test shows that the independent director network and knowledge spillover strengthen the positive effects of moderate dividend payout on green innovation persistence. The Independent Directors’ Network mitigates the negative impact of excessive dividend payout on green innovation persistence. The impact of dividend payout on firms’ green innovation persistence is more pronounced in non-state-owned firms and high-tech firms. Depending on the shareholder type, the inverted U-shaped curve between dividend payment for shareholding managers and green innovation persistence is the steepest. This study provides strong evidence on the impact of dividend payment on green innovation persistence and its role as a mechanism in a more comprehensive way, enriching the literature on corporate green innovation. In addition, it provides Chinese companies with a reference for formulating appropriate dividend policies according to local conditions, strengthening incentives, improving knowledge management capabilities, and further promoting the persistence of green innovation in companies.

Based on the empirical results, this paper formulates the following recommendations: First, companies should fully exploit the governance and incentive function of a moderate dividend payout. By providing cash incentive bonuses to managers and R&D personnel, they can mitigate agency conflicts and improve knowledge management capabilities, which in turn can promote the persistence of green innovation and accelerate the green transformation of the economy. Second, companies also need to limit excessive dividend payouts to curb shareholders’ wealth transfer behavior and prevent the loss of intellectual capital that ultimately hinders their green innovations. Third, companies can further strengthen the supporting effect of moderate dividend payouts on the persistence of green innovation by improving the internal supervisory environment (network of independent directors) and the external knowledge environment (knowledge spillover) to achieve a mutually reinforcing effect.

However, there are some limitations of this paper. First, the research in this paper focuses on the impact of dividend payment on the persistence of green innovations. Its essence is to study the effect of dividend payment as a governance and incentive tool for listed companies to promote the green transformation of the real economy. However, existing studies have found that promoting the sustainable development of the green economy requires not only increasing green innovation performance, but also reducing the risk of green innovation, improving resource efficiency, promoting innovation spillovers, and increasing the business value of the green economy [73,74]. Future research could further investigate whether dividend payments can promote the persistence of green innovations while reducing the risk of green innovations, improving the efficiency of resource use, and achieving effective spillover effects of green innovations within the same region. This means that, on the basis of this paper, the further positive effects of dividend payments in the broader area of the green economy will be further discussed. Secondly, dividend is just one way for shareholders and management of listed companies to receive monetary benefits. Major shareholders and management can also achieve excess returns through trading behavior on the capital market or the intra-group market [75]. How do these excess benefits affect companies’ green innovations? There are fewer studies in this area. Future research could further examine the impact of different forms of benefits, such as excess benefits for large shareholders and private benefits for management, on the persistence of green innovations and also examine whether dividend payments can play an interactive role in this context. Finally, this study uses Chinese listed companies as a research sample, which can to some extent reflect the relationship between dividend payments and green innovation in Asian countries. However, the dividend policies of developed countries such as the United States and the United Kingdom are different from those of China. Research results may differ if relevant data from other countries are adopted as a research sample. Future research could attempt to collect relevant data from multiple countries to examine in depth the differences in the impact of dividend payments on the persistence of green innovations in different countries.