Potential contributions of digital finance to alleviating the ‘low-end lock-in’ dilemma for green innovation in enterprises

As climate change risks intensify worldwide, green technological innovation by enterprises has become a crucial factor affecting the balance between economic development and ecological governance. This paper utilizes data from Chinese A-share listed companies in heavily polluting industries from 2011 to 2021 to investigate the impact and mechanism of the knowledge spillover effect of enterprise digital finance development on the phenomenon of ‘low-end lock-in’ in green innovation. The study finds that digital finance development significantly promotes green innovation in enterprises, with a more pronounced enhancement in high-end green innovation output, thereby mitigating the phenomenon of ‘low-end lock-in’ in green innovation. Mechanism analysis reveals that the development of digital finance in enterprises facilitates high-end green innovation by alleviating financing constraints and enhancing internal control levels through internal and external governance structures. Heterogeneity tests indicate that the promotion effect of digital finance development on high-end green innovation is more pronounced in samples of state-owned enterprises, large and medium-sized enterprises, and enterprises in central and eastern regions. This paper constructs an index of digital finance development for enterprises through text analysis, providing theoretical support for micro-enterprise research on digital finance development and empirical support for the impact of financial development trends on theories of enterprise green innovation.


Introduction
Climate change has become a serious challenge for all countries globally, and the development of green innovation is deeply influencing human lifestyles.According to the latest data from the European Commission's 'GHG Emissions of All World Countries 2023,' the greenhouse gas emissions of the European Union (EU) showed the most significant decline in 2022, decreasing by 27%, achieving a decoupling of greenhouse gas emissions from economic growth.Meanwhile, Russia and the United States decreased by 15.5% and 2.4%, respectively.However, emerging economies such as China and India saw a substantial increase in greenhouse gas emissions, with China and India increasing by 2.85 times and 1.70 times, respectively.As a key driver of economic development and one of the responsible parties for regional environmental pollution, enterprises bear the important responsibility of balancing economic development with environmental governance.However, the level of green technological innovation in enterprises has long been trapped in a dilemma where 'quantity' outweighs 'quality' (Chen et al 2023).For example, China made significant progress in the growth rate of green patent applications from 2010 to 2015 (Wang et al 2019), with the growth rate reaching 50% even in 2016.However, in recent years, the number of patent applications has dropped significantly, with the growth rate falling below 10% in 2019.Looking at the types of patent applications in recent years, the proportion of green utility model patent applications accounts for the majority of total patent applications, indicating a phenomenon of 'low-end lock-in' in enterprise green innovation 1 .The environmental pollution brought about by the 'extensive' economic development of emerging economies has become a key concern for governments (Fujii 2021), thus encouraging enterprises to improve the quality of green innovation is one of the urgent issues for both governments and the academic community to address (Fujii and Managi 2019).
Long-term stable funding is an important prerequisite for enterprises to engage in green innovation.The phenomenon of financing constraints often hinders enterprises' high-end green technological innovation projects.Due to the inherent 'attribute mismatch,' 'sector mismatch,' and 'stage mismatch' in traditional finance, some enterprises face high financial exclusion.Therefore, blockchain, artificial intelligence, and other new-generation information technologies have accelerated breakthroughs, forming a digital financial service model.The development of digital finance provides financial security for enterprises while also bringing about the flow of new information elements (Lin andMa 2022, Ma 2023).
Previous research on the topic of green technological innovation has shown heterogeneous driving factors (Zhou et al 2021).Factors such as Eco-industrial parks (Wu et al 2023), environmental regulation (Kesidou and Wu 2020, Zhang et al 2022a, 2022b), resource constraints (Lin and Ma 2022), and environmental information disclosure (Liu et al 2023, Ren et al 2023) all affect enterprises' willingness and ability to engage in green technological innovation.Related literature has demonstrated that the emergence of digital platforms facilitates the transfer of enterprise knowledge, thereby improving innovation performance (Tian and Hong 2022).Previous studies have focused on external environmental regulations and resource constraints, while research from the perspective of financial service reform is relatively scarce.Currently, digital financial services are increasingly rising, but the environmental impact of digital technology is uncertain (Yang and Wang 2023), and there is a lack of research in the academic community on the factors influencing green innovation from the perspective of knowledge spillover on digital financial service platforms.Although research on the factors influencing enterprise green technological innovation is rich, it is necessary to integrate the development of digital financial services with enterprises' willingness for high-end green technological innovation into a unified research framework.
To address the aforementioned research gap and considering China as a typical representative of emerging economies globally, this paper selects data from listed companies in the heavily polluting industries of China's Shanghai and Shenzhen stock markets from 2011 to 2021.It matches the degree of digital financial development with the number of applications for corporate green innovations to investigate the impact of digital financial development on the level of corporate green innovation.This paper aims to explain the influence of digital financial development on corporate high-end green innovation and the improvement in mitigating the 'lock-in' effect on corporate low-end green innovation from the perspective of knowledge spillovers brought about by digital finance.This paper primarily addresses the following research questions: Can digital finance significantly influence corporate green technological innovation?If significant economic effects exist, is this effect robust?Can the knowledge spillover effect brought about by digital finance promote corporate high-end green innovation?How should China advance the layout and exploration of digital financial development and corporate green innovation in the future?
Compared to existing literature, this paper, for the first time, incorporates the perspective of knowledge spillovers from digital finance into the analysis framework of corporate green technological innovation.Through econometric methods such as the PPML model, it empirically tests the economic effects and impact mechanisms of digital financial development on corporate green innovation, making certain marginal contributions to existing research.Specifically: First, existing studies measure digital finance based on survey questionnaires (Ravikumar et al 2022) and macro perspectives (Liao et al 2022), while this paper utilizes Python technology to identify annual report texts of listed companies, constructing an index of corporate digital financial development to better integrate digital finance theory into the corporate level and provide theoretical reference for the construction of digital financial development databases in other economies.Second, this paper innovatively distinguishes between the low-end and high-end technical capabilities of corporate green innovation and introduces the perspective of knowledge spillovers from digital financial development into the domain of corporate green innovation willingness, enriching the production factor theory in microeconomics and providing reference for the 'extensive development' of other emerging economies.Third, this paper differentiates the characteristics of enterprises, dividing them into different characteristic data samples and further incorporating enterprise heterogeneity characteristics into the theoretical research analysis framework, thereby expanding the research framework of enterprise characteristics in green innovation.
The empirical findings of this study indicate that the better the development level of digital finance in enterprises, the higher the level of green innovation, and the more pronounced the promotion effect on highend green innovation in enterprises.After a series of robustness tests, these conclusions remain unchanged.Furthermore, we found in our research that alleviating financing constraints and enhancing internal control levels are the mechanisms through which digital finance development promotes high-end green innovation.At the same time, in the heterogeneity test results, we found that the promotion effect of digital finance development on high-end green innovation in enterprises is more pronounced in samples of state-owned enterprises, large and medium-sized enterprises, and enterprises in the central and eastern regions.These conclusions provide theoretical basis and empirical reference for the resource guarantee of digital finance development and the output of high-end green innovation in enterprises.
The remaining parts of this paper are as follows: the second part is a literature review, theoretical analysis, and hypothesis formulation; the third part is the research design, including data sources, model design, and variable definition; the fourth part is the empirical research results, including data regression results, robustness tests, and mechanism tests; the fifth part is further examination, specifically theoretical analysis and empirical results of enterprise heterogeneity; the sixth part is the discussion, which discusses the limitations of the data and the innovation of this paper; the seventh part is the conclusion and policy suggestions.
2. Literature review and theoretical analysis 2.1.Literature review 2.1.1.Economic consequences of the development of digital finance Financial development plays a pivotal role in the investment activities of the real economy and enterprises.Effective financial provision can facilitate enterprises in undertaking technological innovation activities (Rao et al 2022).Existing studies predominantly analyze the factors influencing green innovation from the standpoint of green finance (Xu 2023, Zeng et al 2023), with fewer delving into the realm of digital finance.Essentially, digital finance represents a novel financing model emerging from the utilization of information technology to address the high financing costs and risks associated with traditional finance.Research on digital finance can be categorized into three main areas.Firstly, the measurement of digital finance is paramount.Scholars in this domain construct a regional digital finance development index based on the depth of digital financial services and the operational environment of a city (Liao et al 2022).Other studies assess sample enterprises' digital financial development levels through questionnaires (Ravikumar et al 2022, Thathsarani andWei 2022).Secondly, there is the issue of how digital finance impacts the development of macroeconomics and traditional finance.For instance, scholars have examined the role of digital finance in promoting regional green innovation (Shao and Chen 2023), industrial green transformation (Zhong et al 2023), and the inclusivity and growth of the financial industry (Ozili 2018).Thirdly, the influence of digital finance on the micro-behaviors of enterprises is significant.Relevant academic research primarily elaborates on two aspects: resource acquisition and information transfer.Firstly, the development of digital finance employs internet technology to organize information into a network structure, eliminating information 'silos,' fostering cross-validation of information, and thereby enhancing the level of information symmetry among enterprises (Kong et al 2022).This, in turn, boosts the efficiency of enterprises in acquiring capital resources, alleviating financing constraints, and reducing the risk of financial defaults.Secondly, digital finance utilizes internet interaction technology to achieve enterprise information transparency, minimizing the level of information asymmetry among enterprises, fostering informal communication within enterprises, and enhancing organizational operational efficiency (Chang et al 2023).Lastly, digital finance facilitates effective information supervision, enabling the realization of an intelligent supervision model for enterprises and effectively overseeing enterprises' green innovation R&D activities.The application of this financing approach allows for real-time assessment of enterprises' risk-taking capabilities through big data (Ji et al 2022).By regulating investment levels, it serves the functions of preevaluation, mid-event feedback, and ex-post supervision for enterprises' green innovation investments.
Moreover, the literature has also focused on the knowledge spillover effect of the development of the digital economy.In the initial stages, Michael Polanyi proposed the classification of explicit knowledge and tacit knowledge in 1958.Explicit knowledge refers to standardized knowledge that can be fully and clearly encoded, stored, and disseminated, and can be spread across time and space with the assistance of the internet.On the other hand, tacit knowledge is highly context-dependent and challenging to articulate beyond specific contexts and situations (Hao et al 2017), making it difficult to disseminate through digital channels.Numerous studies suggest that digital information technology reduces the barriers of space and time for knowledge dissemination, as well as the costs associated with knowledge innovation and acquisition, thus effectively facilitating knowledge diffusion (Antonelli 2017, Colombelli et al 2023).The knowledge spillover generated by digital finance mainly operates through two transmission mechanisms: The first effect is the aggregation effect.In the context of imperfect competition, the advancement of digital finance reduces the costs associated with network information collection and transmission, thereby amplifying the aggregation of competitive advantages in green innovation within enterprise spatial centers.Additionally, the evolution of digital finance enhances digital infrastructure for enterprises, attracting the spatial concentration of production resource elements such as technology, talent, and data, driving synergistic aggregation between high-tech enterprises and China's hidden champions, ultimately forming a knowledgeintensive innovation network (Li et al 2022).Furthermore, digital finance utilizes data resources to facilitate capital dispersal and connect industrial advantages across time and space.This large-scale knowledge network effect fosters an environment conducive to the green and efficient innovation of enterprises, accelerating the transfer and absorption of tacit knowledge.
The second effect is the diffusion effect.Enabled by internet technology, digital finance transcends temporal and spatial constraints, leading to the spillover of R&D elements from central cities to peripheral cities, in line with the law of diminishing marginal returns to capital in Solow's economic growth theory, thereby establishing a spatial spillover effect (Yang et al 2023).This process contributes to an increased presence of green innovation industries in peripheral cities, ultimately promoting a more balanced distribution of regional green innovation and indirectly stimulating the green innovation output of peripheral cities (Li et al 2023a(Li et al , 2023b)).The knowledge diffusion effect facilitated by digital technology enables enterprises to embrace advanced environmental protection concepts, aiding governmental observation of informal environmental regulations influenced by public opinion (Zhang and Huang 2023).Furthermore, it facilitates the transformation of enterprise green governance into standardized and precise control, enabling the absorption of advanced technology in green environmental governance and enhancing productivity and scientific pollution control measures.

Influencing factors of enterprise green innovation
In the exploration of factors influencing enterprise green innovation, existing literature primarily delves into environmental regulation, organizational structure, corporate governance, and governmental influences.Regarding environmental regulation, scholars diverge on its impact.Some argue that stringent regulations escalate enterprises' legitimization costs, thereby diminishing innovation investment and impeding green innovation (Xepapadeas and Zeeuw 1999, Chen et al 2022, Fang and Shao 2022).Conversely, others contend that environmental regulations surmount organizational inertia and optimize structures, thus fostering green innovation (Xepapadeas and Zeeuw 1999, Chen et al 2022, Fang and Shao 2022).In terms of organizational structure and corporate governance, studies indicate that enhanced corporate information transparency and streamlined organizational management significantly bolster enterprises' capacity and willingness for green innovation (Kong et al 2022, Tan et al 2022).External oversight and management further enhance enterprises' independent motivation for green innovation and facilitate independent R&D endeavors in green innovation (Lu et al 2023).Governmental factors also play a crucial role.Relevant policies can supervise enterprises' focus on environmental performance to prevent the unfettered expansion of the existing economy, thereby compelling enterprises to engage in green R&D and innovation and averting the 'crowding out effect' on green innovation (Yi et al 2019, Sun et al 2022, Zhang et al 2022a, 2022b).According to resource dependence theory, external financing is imperative for enterprise green innovation (Yang et al 2022, Tang et al 2023).Given the high risk inherent in high-end green technology innovation, the choice of financing methods becomes pivotal.Digital finance, leveraging big data technology, bridges the bank-enterprise gap, rectifies enterprise information asymmetry, alleviates financing challenges, and propels breakthroughs in green innovation (Rao et al 2022).Furthermore, digital finance's development enables the establishment of corporate financial risk early warning mechanisms through computer algorithms.This facilitates effective risk assessment via data streams, augments resource utilization, refines resource allocation, and supplements conventional finance methods.Consequently, it monitors enterprises' output in green high-end innovation.

Literature review
Based on the relevant literature, several key points emerge regarding research on digital finance.First, there is a lack of unified measurement standards for digital finance indicators in China.Although the widely used Peking University digital finance development index exists, it struggles to align with enterprise data.Consequently, employing a mix of macroeconomic and microeconomic data can introduce significant errors into regression results.Thus, there is a pressing need in academic research for an index that can be correlated with enterprise data to accurately gauge the digital financial development of enterprises.Secondly, existing research on digital finance intersects with the study of green innovation in finance and enterprises, as exemplified by the works of Rao et al (2022) and Chen et al (2023).However, this paper diverges from these studies in several ways.While Rao et al (2022) focuses on the enterprise's green innovation capability through the number of green patents granted, they pay less attention to the enterprise's independent willingness for green innovation.This study contends that evaluating the volume of green patent applications provides a more comprehensive assessment of an enterprise's innovation motivation and potential, as well as highlighting the challenge of 'low-end lock-in' for green innovation.On the other hand, Chen et al (2023) investigates the impact of digital finance on green innovation outputs from a spatial econometrics perspective, emphasizing the spatial spillover of digital finance as the primary research angle.This study diverges from their methodologies and perspectives.
This paper presents an innovative perspective by examining the information potential of digital finance from the angle of knowledge spillover, thereby providing theoretical supplements to the study of digital finance.
Lastly, while green innovation is predominantly studied from the standpoint of overall green technology levels, there remains a dearth of research on differentiating high-end green innovation from low-end green innovation.Additionally, there is a lack of enterprise-specific studies on the influence of digital finance on the development of high-end green innovation.In summary, this paper aims to address these research gaps in the fields of digital finance and enterprise green innovation, offering a deeper exploration of relevant theoretical aspects.

Theoretical analysis
Low-carbon development is not only a goal for sustainable development of enterprises but also a vision for governments to achieve economic transformation.Therefore, environmental pressures on enterprises necessitate considering the cost of environmental governance in the production process (Lu et al 2023, Chen et al 2024).Green technological innovation in enterprises aims at energy conservation and emission reduction, which can help reduce pollutant emissions and alleviate environmental regulatory pressures on enterprises.To better elucidate the decision-making process of enterprise green innovation under the backdrop of digital finance, this paper employs a two-stage decision-making model for theoretical derivation.
Assuming the enterprise is a heavy polluting enterprise, the cost of environmental governance will have a significant and long-term impact on the profit of enterprise production and operation.Therefore, the enterprise needs to consider whether to engage in high-end green innovation.Before the application of green technological innovation in the final product, the enterprise's research and development department will design intermediate products of green innovation technology, which will continuously improve in quality with the continuous updating and iteration of technology.We primarily discuss the decision-making process of enterprises on whether to engage in high-end green technological research and development under the influence of knowledge spillover effects on digital finance platforms.This paper adopts a two-stage game to elucidate the decisionmaking process and uses backward induction to determine the optimal strategy.We first focus on the production end.

Production end
In this stage, the quality of the green research and development products (G) provided by the research and development department remains stable, and it is expected that they can be successfully applied to energy conservation and emission reduction in the enterprise.Specifically: Energy-saving and emission-reduction technologies need to continuously innovate and improve in quality in line with the advancement of green innovation capabilities and the requirements of government environmental policies.Therefore, 'X' is a quality adjustment quantity of 'G'.
However, whether the emission reduction effect can meet the standards depends on the extent of enterprises' investment in green innovation.Additionally, 'X' as a well-developed green innovative product can be patented and licensed for additional revenue.Furthermore, if a company can have better research and development output in green innovation, producing more green patents, it will have higher financing efficiency when using digital financial services.The financing efficiency of digital finance is denoted as 'h', that is, 'h F X , ( )'. 'F' represents the company's own production and operation status.According to endogenous growth theory, the knowledge stock of enterprises increases through the increase in capital stock input.When companies use digital financial platforms for financing, it promotes information sharing between enterprises and between enterprises and investors.In these interactions, companies may obtain information about green technology innovation, including emerging environmental protection technologies and sustainable development practices.The impact of obtaining information about the level of green technology in the same industry on the company's own investment in green innovation technology is called knowledge spillover effect.Therefore, the knowledge spillover in the research and development stage is denoted as The following text will take a heavily polluting enterprise as an example, assuming that the enterprise operates in a perfectly competitive market, with a Cobb-Douglas production function: Where L represents labor input and K represents capital input, with a b + < 1 as the total factor productivity.We assume in enterprise B that the marginal input of labor is the same across all departments, but the capital input is divided into productive capital input and green technology input.The proportion of funds allocated to green technology investment in the enterprise is denoted as s, and < < s 0 1 denotes the total capital input used for production by the enterprise, wheres K 1 ( ) represents the capital input used for productive purposes.During the production process, enterprises need to consider both economic benefits and environmental policy constraints imposed by the government.Government environmental policies are exogenous to the firm.
If the government places greater emphasis on environmental utility, it imposes more stringent environmental policies, resulting in higher environmental compliance costs for the firm.To clarify, we distinguish between the revenue and costs of the firm.

The firm's revenue
To focus more on studying the green technology innovation of heavily polluting enterprises, this paper mainly considers the revenue of the enterprise derived from the sales of finished products and the income from transferring green technology patents.Other non-operating income of the enterprise is not taken into account.Therefore, the revenue function of the enterprise is: Among these, R represents the revenue of the enterprise, p F is the equilibrium price of the enterprise's finished products, and p G is the equilibrium price for the transfer of green patents by the enterprise.

The firm's cost
The main costs for enterprises lie in production costs and environmental regulation costs; therefore, the cost function for enterprises is: ) represents the production costs of the enterprise,and w is the marginal cost of labor, while r represents the marginal cost of capital input.⋅ E ( ) is the environmental cost function for the enterprise, which has four methods of responding to environmental regulations.The costs of the 4 options are as follows: Green ) and G is a function of research investment (sK), knowledge spillover (m) and the environmental policy (t); Purchasing trading permits: ) where s is the equilibrium price of carbon trading.It is a function of production and environmental policy; Purchasing other green patent or facilities: ) where N is the number of facilities the firm has to buy; Damage cost: D t , ( ) the total fine of excess emissions, which is a function of environmental policy (t).Thus, from (3) and (4), it is known that the profit objective function for enterprise i is: If the environmental costs are lower than the profits obtained from excess emissions, enterprises will still opt to emit pollutants rather than pursue green technology innovation.Therefore, enterprises need to ensure that the marginal environmental cost of green innovation is minimized, then: Therefore, combining the condition in (7), the optimal green technology innovation input, sK, can be derived from (6) as follows: Based on the calculations above, it indicates that digital finance has a positive effect on enterprises' green technology investments.Therefore, Hypothesis 1 is proposed: H1 Holding other conditions constant, digital finance contributes to enhancing the level of green innovation in enterprises.

Research and development (R&D) end
The output of the enterprise's R&D department is intermediate goods in green innovation (G).Due to changes in the economic environment and government environmental policies, green innovation also needs continuous updating and iteration.
Assuming that the green innovation of the department needs to achieve the highest quality q k ( ) for the current period, where k is the current green technology level of the firm.The probability of successful green research and development (R&D) p k Z , ( )depends on the total expenditure ⋅ Z ( ) of the department's green R&D and the green innovation technology level k of the enterprise, following a Poisson Process.This implies: Generally, the R&D department's R&D investment is positively correlated with the probability of R&D success, denoted as If the current technology level k of enterprise i is lower than that of its peers, indicating a lower value of k, the early probability of innovation success is higher due to the 'knowledge spillover' utility of the data platform.However, if the industry's technology level is already high, the difficulty of innovation will increase, and the probability of success will decrease.The relationship between k and ⋅ p( ) is quadratic.It is important to note that the total R&D expenditure Z is different from the total green technology investment sK in the second step.sK tends to be the environmental compliance cost of green innovation products (X) that are more in line with production and technically mature, while Z tends to be the total R&D expenditure of green innovation (G) influenced by the enterprise's existing environmental protection technology.
Assuming m is the knowledge spillover function, it is a function of the current green technology level k of the enterprise.The probability of successful green R&D by the enterprise is determined by the following equation: Herein, the ⋅ Z ( ) of this stage is determined by the current industry's technology level and the 'knowledge spillover' effect m .
( ) Furthermore, due to This indicates that the 'knowledge spillover' effect of digital finance platforms improves the efficiency of enterprise green research and development (R&D) investment, thereby increasing the probability of R&D success.
Specifically, during the financing process, enterprises can use digital finance platforms to understand the degree of correlation between green technology and capital (e.g., what level of achievement makes it easier for enterprises to obtain financing approval).In addition, digital finance platforms can provide specialized green financial products and services.Through digital finance financing, enterprises can more easily access relevant information and resources on green technology innovation, thereby understanding the current level of green technology in the same industry (e.g., which technology is more effective in energy conservation and emission reduction).This reduces the 'innovation consumption' of enterprises in technological innovation, specifically manifested as an increase in the probability of success in green innovation.m k ( ) represent the impact of the current level of technology on the probability of success, which is a quadratic term.If an enterprise's current level of green technology is low, then m can increase the efficiency of the enterprise's R&D.If an enterprise's level of green technology is already very high, then higher quality green innovation poses a challenge to the enterprise.Define its relationship with output quality as: In which, e represents the parameter for research and development investment, and e > 0. By substituting (11) into (9), the relationship between the total research and development investment of the enterprise and the quality of the output of green intermediate products (G) can be determined as follows: By substituting (12) into (8), we have: This indicates that the 'knowledge spillover' utility of digital finance contributes to reducing the consumption of high-quality green innovation by enterprises, thereby enhancing the research and development (R&D) efficiency and willingness of enterprises for high-end green innovation.Therefore, we propose Hypothesis 2: H2 : Holding other conditions constant, the development of digital finance contributes to enhancing the level of high-end green innovation in enterprises.

Sample selection and data sources
This study focuses on A-share listed companies operating in heavy pollution industries on the Shanghai and Shenzhen Stock Exchanges in China from 2011 to 2021.The initial sample selection process is outlined as follows: Firstly, companies belonging to 16 categories of heavy polluting industries are identified, including thermal power, steel, cement, electrolytic aluminum, coal, metallurgy, chemical industry, petrochemicals, building materials, papermaking, brewing, pharmaceuticals, fermentation, textile, tanning, and mining.Secondly, samples with missing or abnormal data are excluded.Thirdly, companies labeled as ST or * ST are also removed from the sample.Fourthly, to mitigate the impact of outliers, continuous variables are rounded to the nearest integer.Subsequently, a total of 6508 annual industry observations are obtained.Financial data used in the analysis are sourced from the CNRDS and CSMAR databases.

Variable measurement 3.2.1. Dependent variable: green innovation
In the selection of indicators for measuring green innovation in enterprises, existing studies typically quantify them through the lenses of innovation input and output.This study, however, focuses on the perspective of green innovation output by using independently applied green patents as quantitative indicators of enterprise green innovation.It's worth noting that patents encompass inventions, utility models, and designs.Green invention patents serve as a tangible reflection of substantive green innovation within enterprises and possess high value, whereas utility model green invention patents represent strategic green innovation and hold lower value.Furthermore, in accordance with relevant laws and regulations, the application process for green invention patents involves both formal and substantive examination, whereas green patents for utility models only require formal examination.
Therefore, considering the value of patents and the level of scrutiny in patent applications, this article adopts enterprise green invention patents as a metric for high-end green innovation, and utility model patents as a metric for low-end green innovation.To assess the comprehensive impact of digital finance development on green innovation and to discern differences in impact between high-end and low-end green innovation, the overall level of green innovation is utilized to evaluate enterprise green innovation development.The specific measurement method involves calculating the total number of green invention patent applications and green utility model patent applications submitted by the enterprise.

Independent variable: enterprise digital financial development
Describing the level of digital finance development at the enterprise level accurately is a challenging task due to the systemic nature of this process.Existing studies have predominantly approached this from a macro perspective, utilizing the Peking University Digital Finance Development Index to measure interregional digital finance development levels (Liao et al 2022), while others have relied on questionnaire-based surveys for measurement (Ravikumar et al 2022, Thathsarani andWei 2022).However, these methods have their limitations.For instance, the Peking University Digital Finance Development Index relies solely on personal data from Alipay, neglecting data from other digital payment platforms.Moreover, the index, being a macrolevel regional variable, may not be appropriately matched and regressed with micro-enterprise data, leading to dimensional bias in regression analysis.On the other hand, the questionnaire method, targeting individual enterprises, lacks the ability to generate large datasets, resulting in experimental results that exhibit regional and individual discrepancies and, therefore, lack universal applicability.
Given these limitations, this paper proposes an alternative approach.We construct a relatively comprehensive semantic dictionary of enterprise digital finance development based on the semantic expression of national policies related to digital finance.Subsequently, we employ a machine learning-based text analysis method to develop a comprehensive indicator reflecting the degree of digital finance development among A-share listed companies in China.The specific steps involved are outlined below: Step 1: Constructing the Lexicon of Enterprise Digital Finance Terminology.In response to the absence of specialized lexicons in the field of digital finance, this study addressed this gap by constructing a lexicon of terms based on the semantic framework of national policies.Drawing upon relevant studies (Razzaq and Yang 2023), keywords related to enterprise digital finance were extracted from 69 selected official policies and announcements retrieved from the websites of central, provincial, and municipal governments, as well as the Ministry of Industry and Information Technology.After undergoing word segmentation using Python, relevant vocabulary concerning the development of enterprise digital finance, appearing at least five times, was selected.Following manual verification and supplementation, a final selection of 355 words related to the development of enterprise digital finance was compiled, forming the lexicon utilized in this paper.
Step 2: Conducting Text Analysis on Relevant sections of Annual Reports.The selected 355 lexicon items were integrated into the 'jieba' Chinese word segmentation library within the Python software package.Subsequently, employing a machine learning approach, text comparison analysis was conducted on the 'Management Discussion and Analysis' (MD&A) section of the annual reports of listed companies.The frequency of occurrence of the 355 words in the annual reports of each company for each year was calculated.
Step 3: Constructing the Index for the Level of Enterprise Digital Finance Development.In addressing the variances in the length of the MD&A section within annual reports across different companies, this study referenced pertinent literature (Luo et al 2023) and measured the micro-level development of enterprise digital finance by computing the sum of frequencies of relevant words divided by the length of the MD&A section in the annual reports.This methodology serves to mitigate measurement errors stemming from discrepancies in the lengths of MD&A sections in annual reports.The rationale behind formulating this index lies in the premise that if an enterprise adopts digital finance practices, it is more inclined to feature such information in its annual report.In order to facilitate the interpretation of regression coefficients in various regression models, this article standardized the independent variables.Consequently, a higher value of the index of enterprise digital finance signifies a more advanced level of development in enterprise digital finance.

Financing constraint
Referring to the study of Hadlock and Pierce (2010), this paper establishes a model to measure the degree of corporate financing constraints: 1 14 where size represents the asset size of the enterprise, measured by the natural logarithm of total assets; lev represents the financial leverage ratio of the enterprise, which is measured by the asset-liability ratio = (total liabilities/total assets); CashDiv represents the cash dividends paid by the company in the same year; and MB represents the enterprise price-to-book ratio, which is measured by market value/book value.NWC represents net working capital, and the measurement indicators are working capital-monetary funds-short-term investment; EBIT represents profit before interest and taxes; and TA represents total assets.In the first step, three variables -company size, company age, and cash dividend payment rate -were standardized on an annual basis.The listed companies were then sorted in ascending order according to the mean values of the standardized variables.The upper and lower tertiles were used as financing constraints.The threshold for the financing constraint dummy variable OUFC was determined.Listed companies above the 66th percentile were categorized as the low financing constraint group (OUFC = 0), while those below the 33rd percentile were categorized as the high financing constraint group (OUFC = 1).
In the second step, logit regression was performed on Model (1) to determine the probability P of the enterprise's financing constraint for each year, defined as the financing constraint index FC, with values ranging between 0 and 1.A higher FC value indicates a more severe financing constraint for the enterprise.

Level of internal control
Internal control serves as a crucial indicator of an enterprise's organizational capacity.The five elements comprising internal control offer a means to evaluate the internal operational efficiency of an enterprise.Existing literature commonly employs external rating indices to gauge the level of internal control within an enterprise.In this study, we adopt the Dibo variant of the internal control index to assess the level of internal control in enterprises.A higher value of the internal control index signifies the implementation of more stringent internal control measures by the company.

Control variables
Building upon the study conducted by Rao et al (2022), this paper has identified several control variables, including the asset-liability ratio, net profit rate of total assets, cash flow ratio, proportion of fixed assets, growth rate of operating income, shareholding ratio of the largest shareholder, establishment years of the company, shareholding ratio of management, management expense ratio, and capital occupation of major shareholders.
Specifically, a higher asset-liability ratio implies that a company is more cautious in its use of funds, prompting it to seek more efficient asset utilization and thereby driving green innovation.However, a high assetliability ratio can also subject the company to greater financial risk pressure, limiting its investment in green innovation.An increase in the net profit margin of total assets signifies that a company can more efficiently utilize assets, allowing it to have more cash flow available for green innovation projects.Similarly, stable cash flow may enable a company to invest more in green innovation.A higher proportion of fixed assets implies that a company has a certain production base and is more conducive to green technology innovation.Furthermore, a high growth rate in operating income indicates that a company has greater profit potential, but an excessively high growth rate may lead management to allocate resources to projects with greater profitability, potentially neglecting investments in green innovation.
As a company′s operational tenure increases, it may encounter resistance to green innovation due to traditional innovation management and inertia.However, in terms of the relationship between operational tenure and cash flow, mature companies will have more cash flow and be willing to try new green technologies.The impact of major shareholder funds utilization and management expense ratio on corporate green innovation is similar.Regarding the proportion of shares held by shareholders and management, the willingness of a company to engage in green innovation depends on the resources and motivations of shareholders and management, making it multidimensional and variable for corporate green innovation projects.In summary, all of these variables have an impact on green innovation, making it necessary to incorporate them into regression models.
The definitions for these variables are shown in table 1:

Empirical models
To better handle zero-inflated data in corporate green innovation, this study employs PPML for estimation and testing.Model is as follows: where Green represents enterprise green innovation, including overall green innovation, high-end green innovation, and low-end green innovation; DF represents enterprise digital financial development; Controls represents the model control variables; β i represents fixed effects for years, and ω i represents industry fixed effects.In the regression model, clustering at the firm level is employed to better control for the correlation of error terms.

Analysis of the empirical research
4.1.Descriptive statistics for the variables Table 2 presents the descriptive statistics of the main variables.As shown in table 2, the minimum value of the overall green innovation level of enterprises (Green_T) is 0, the maximum value is 198.00, and the mean is 1.11.This indicates that the overall green innovation level of enterprises is relatively low, with significant fluctuations among them.The average value of high-end green innovation (Green_R) is 0.64, with a median of 0, suggesting that the overall level of high-end green innovation among Chinese listed companies is low.Meanwhile, the level of low-end green innovation (Green_C) is also low, indicating significant room for improvement in both highend and low-end green innovation levels in China.The mean value of digital finance development level (DF) (0.79) exceeds the median (0.70), indicating an overall trend of high development in digital finance among listed companies, consistent with the rapid development trend of China's digital economy.The remaining variables are similar to existing literature and are not further elaborated here.

Analysis of the main regression results
Based on empirical statistics, the model's VIF value is 2.24, far below the threshold of 10, indicating the absence of multicollinearity issues in the regression model.Table 3 presents the regression results.Controlling for year and industry effects and clustering at the enterprise level, the regression results are considered reliable.The results show that, among the three columns in the table, the coefficient for enterprise digital finance development (DF) is significantly positive at the 1% level for Green_T, Green_R, and Green_C, with regression coefficients of 0.205, 0.221, and 0.153, respectively.This indicates that the level of enterprise digital finance development indeed promotes the output of green innovation within the enterprise, thus confirming hypothesis H1.By comparing the correlation coefficients of the three variables, it can be observed that digital finance development significantly promotes overall enterprise green innovation, with a more pronounced effect on high-end green innovation compared to low-end green innovation, thus validating that enterprise digital finance development can alleviate the phenomenon of 'low-end lock-in' of green innovation.Therefore, hypothesis 2 is validated.Digital finance contributes to enhancing the intensity and effectiveness of financial support for green innovation.Due to the difficulty in quickly realizing the environmental benefits of green innovation, coupled with constraints related to innovation risks, enterprises' green innovation activities face challenges in obtaining effective financing.Digital finance exhibits typical inclusive characteristics, naturally possessing advantages such as low thresholds and high efficiency, which facilitate the precise matching of financing needs between supply and demand parties.Additionally, under the inducement of digital finance, social capital quickly converges towards green innovation projects with high potential and value, thereby enhancing the targeted financial support for green innovation and exerting a stronger influence on high-value green high-end innovation.Furthermore, based on the theory of asymmetric information, digital finance can break the temporal and spatial constraints of green technology innovation information transmission, accelerating the movement of financial service demands towards heavily polluting enterprises and compensating for the information asymmetry in financial markets.Consequently, digital finance can shorten the ineffective waiting time for financing of heavily polluting enterprises, invisibly enhancing the enthusiasm of all parties involved in investment and financing for green innovation, ultimately promoting enterprise green innovation.

Robustness test 4.3.1. PSM
This study employs propensity score matching (PSM) to address endogeneity issues.Using a 1:1 nearest neighbor matching method, it was found that the standardized differences of all control variables were within 5% after matching, indicating no significant differences in control variables between the experimental and control groups after matching.After matching, regression was conducted on the matched samples according to the regression model, still controlling for the same control variables, industry and year fixed effects, and clustering at the company level.The results in table 4 show that enterprise digital finance development (DF) remains significantly positive at least at the 10% level, confirming the robustness of the research findings and conclusions.

One period lag
Given the time required for enterprises to develop green innovations, there is also a necessity for enterprises to allow for a period of adjustment in their digital financial financing models.Lagged explanatory variables were employed to validate our findings.In this study, the independent variable (DF) and its corresponding control variables were all lagged by the initial period.The findings, summarized in table 5, consistently indicated across three columns that the lagged DF variables were all significantly positive at the 1% level.Consequently, this validates the primary conclusions drawn in this paper, thus establishing the robustness and reliability of our findings which remain consistent with the baseline regression results.

Instrumental variables
To address the potential endogeneity issue between the level of enterprise digitalization and green innovation, this study employs the instrumental variable method.Theoretically, the advancement of enterprise digital finance is expected to mitigate financing constraints and bolster internal control mechanisms within enterprises, consequently fostering greater levels of green innovation.Conversely, the theoretical premise of reverse causation suggests that the development of corporate green innovation does not directly spur the digital financial advancement of enterprises.Therefore, the notion of reverse causation is not supported.However, considering the potential presence of omitted variables and associated errors, as well as the significant correlation between the selection of instrumental variables and the initial requirements of digital finance, the cross-multiplicative term (referred to as 'Tool') between the Internet penetration rate and the development level of digital finance one period lag is chosen as the instrumental variable.The results of the instrumental variable method for the three dependent variables are presented in the table.Additionally, the subsequent findings indicate that in the first stage (refer to Column (1) of table 6), the Tool exhibits a statistically significant positive impact at the 1% level.In the second stage (refer to Columns (2), (3), and (4) of table 6), the variable DF is statistically significant at the 1% level.Therefore, the second-stage regression results are deemed significant and pass the test, providing evidence that the conclusions drawn in this paper are reliable.Specifically, the level of development of an enterprise's digital finance is found to positively influence the enterprise's green innovation.

Add control variable
To mitigate the potential instability of results stemming from omitted variables, this study considers the inclusion of several macroscopic control variables to reaffirm the existing conclusions.Specifically, this paper incorporates per capita GDP (Rgdp), industrial structure, and per capita R&D intensity (RD) to reassess the findings.The rationale behind this approach lies in recognizing that within the macroeconomic context, the level of economic development may influence enterprise capital operations, the industrial structure may impact the stability of the enterprise's supply chain, and regional per capita R&D intensity may affect the R&D vitality of enterprises within the region.Consequently, each variable has the potential to influence the operation of the model.By incorporating relevant control variables, this paper seeks to further scrutinize the empirical results.
The results, as presented in table 7, across the three columns indicate that DF remains significantly positive at the 1% level.As a result, the main conclusion of this paper is reaffirmed, demonstrating robustness and reliability consistent with the baseline regression results.

Mechanism path test
This paper employs the method of intermediary effect analysis to study the mechanism of the impact of digital finance on high-end green innovation in enterprises.Because the chosen intermediary variables (financial constraints, internal control) have a clear and intuitive causal relationship with the explained variable, high-end green innovation, the focus is primarily on examining the influence of the explanatory variable (DF) on the intermediary variables.As all intermediary variables are continuous variables, this paper sets up a bidirectional fixed-effect regression model to test the intermediary effect while clustering at the enterprise level.The results are presented in table 8.
(1) Stable cash flow is crucial for enterprise green innovation.Previous research has indicated that financial constraints can hinder the development of green innovation in enterprises, and the intermediary effect analysis results of digital finance alleviating financial constraints are shown in the first column of table 8. Empirical results demonstrate that the regression coefficient of the explanatory variable DF on the enterprises.This impact mechanism is manifested as follows: If the internal control construction of an enterprise is weak, it will exacerbate the imbalance of power and hinder the sustainable development of the company.Digital finance, through information science and technology, accelerates the circulation and transformation speed of information in the market, enhances the efficiency of information allocation and transparency, thereby strengthening the efficient transmission of internal governance information within the company, and avoiding management hindering self-interested behaviors that impede corporate green innovation.Furthermore, effective internal control can further reduce the management costs of enterprises, improve the level of cash flow, and provide more support for high-end green innovation in enterprises.Digital finance can integrate various types of information from enterprises, conduct quantitative risk predictions, correct corporate management behaviors, shape a power-balanced ecosystem, and thereby promote the improvement of the level of internal control in enterprises.In summary, the development of digital finance is conducive to strengthening enterprise information management and risk control, and improving the level of internal control in enterprises.

Heterogeneity test based on property rights
In this study, the enterprise samples are divided into samples of state-owned enterprises (Panel A) and nonstate-owned enterprises (Panel B).The empirical results in table 9 show that in state-owned enterprises (Panel A), the development of digital finance (DF) significantly promotes green innovation in enterprises at the 1% level.Furthermore, upon closer examination of the empirical results coefficients after distinguishing between different levels of green innovation, it is observed that the promotion effect of high-end green innovation level (Green_R) is significantly higher than that of low-end innovation level (Green_C).One possible explanation is that, in comparison to private enterprises, state-owned enterprises are more inclined to undertake environmental protection tasks to meet the legitimacy requirements of environmental compliance.Conversely, private enterprises may opt for easily replicable and low-cost technological solutions to fulfill environmental protection obligations.Enterprises acquire knowledge through two main channels: internal knowledge reserves and external knowledge acquisition.Digital finance facilitates frequent knowledge interaction, resource exchange, and business transactions among enterprises through data transmission.This fosters a knowledge spillover effect, generating a wealth of diverse and complementary knowledge among enterprises.Such an environment helps mitigate uncertainty surrounding green innovation for enterprises and significantly boosts high-end green innovation.In contrast, state-owned enterprises possess ample knowledge reserves and stocks, making it easier to generate surplus knowledge resources through data analysis and enhance technology spillover.State-owned enterprises are more inclined to respond to government requirements from the perspective of green innovation knowledge spillover.Consequently, in the sample of state-owned enterprises, the impact of digital finance on knowledge absorption and transformation through the knowledge spillover effect is more pronounced.

Heterogeneity analysis by region
In this study, based on the provinces where the enterprises are located, companies are divided into two samples according to the standards set by the Ministry of Industry and Information Technology of China: Central and Eastern regions, and Western region.The regional heterogeneity of the impact of digital financial development on green innovation is analyzed by comparing the effects of digital financial development in the two regions.As shown in table 10, for companies in the Central and Eastern regions (Panel C), digital financial development can significantly promote the level of green innovation in enterprises, with the independent variable (DF) being positively significant at the 1% level.However, for companies in the Western region (Panel D), it cannot be proven that digital financial development promotes the level of green innovation in enterprises, as the independent variable (DF) is not statistically significant.The coefficient of the independent variable (DF) in the empirical results indicates that the promotion effect of digital financial development on high-end green innovation capabilities is significantly greater than that on low-end green innovation capabilities.
Possible reasons include: the extent of green R&D innovation in enterprises largely depends on the geographical location of the enterprise.The more developed and accessible the urban economy where the enterprise is located, the more likely it is to engage in innovative behavior.The more complete the digital infrastructure in the city where the enterprise is located, the more likely it is to attract digital talent for data absorption and reprocessing, combined with the enterprise's own green innovation capabilities for technological upgrading and transformation.Therefore, enterprises in provinces with well-established digital infrastructure are more likely to effectively utilize knowledge spillovers from data.Compared to eastern coastal provinces, where digital facilities were constructed earlier and the level of digital economy is higher, the probability of developing digital financial models in enterprises is higher, thus leading to more apparent knowledge spillover effects.In contrast, although the digital infrastructure in central provinces is not yet fully developed, their unique geographical and transportation advantages, along with the industrial transfer from developed eastern provinces, result in more pronounced knowledge spillover effects from digital financial development in enterprises in central provinces.Enterprises in central provinces are at the center of innovation network interaction in various regions, thus acquiring relatively comprehensive technological knowledge.Additionally, enterprises in central provinces receive more tacit technological knowledge through the transfer of technology industries from eastern provinces, forming an overall advantage in industrial support.Therefore, they can effectively utilize data elements for knowledge utilization.On the other hand, in western provinces, there is considerable room for improvement in digital infrastructure, and both green competitive advantages and transportation costs pose significant obstacles, leading to a lag in the absorptive capacity of knowledge spillovers.In summary, enterprises in central and eastern provinces show better absorption of knowledge spillover effects from digital financial development, while there is no clear evidence of this promoting effect in western provinces.

Heterogeneity analysis by enterprise size
This study distinguishes samples based on enterprise size (natural logarithm of total annual assets).Due to the larger size of heavily polluting enterprises, the sample is divided into two groups based on the median of enterprise size, with samples larger than the 50th percentile defined as large companies (Panel E), and the rest defined as small and medium-sized enterprises (Panel F).The empirical results, as demonstrated in table 11, show that in the sample of large enterprises (Panel E), digital finance (DF) significantly promotes green innovation across all enterprises.Conversely, in small and medium-sized enterprises(Panel F), digital finance (DF) only promotes high-end green innovation.Comparison of coefficients reveals that in the sample of large enterprises (Panel E), the promotion effect of digital finance on high-end green innovation is significantly higher than that on low-end green innovation.The potential explanation lies in the fact that large enterprises possess a distinct advantage in terms of scale, coupled with standardized processing procedures that facilitate the absorption of knowledge spillover effects stemming from data.Notably, the absorptive capacity for knowledge spillover in large enterprises surpasses that of small and medium-sized enterprises (SMEs) significantly.Furthermore, owing to their involvement in a greater number of green research and development (R&D) initiatives, bolstered absorptive capacity, and enhanced foundational research capabilities, large enterprises are better positioned to extract knowledge spillover benefits from the data-centric advantage offered by digital finance.Conversely, SMEs, due to their substantial technology gap compared to large enterprises, allocate less focus towards external knowledge absorption.While SMEs can also reap the rewards of knowledge spillover, they lack the scale advantage enjoyed by large enterprises.Additionally, SMEs exhibit less diversity compared to their larger counterparts, resulting in a less varied array of technology designs.Disparities in firm size inevitably translate into varying degrees of knowledge spillover.Limited access to financing channels, fund constraints, and regulatory oversight within the realm of digital finance present challenges for SMEs, prompting them to prioritize high-end innovation as a means to swiftly penetrate the market and thereby attain a competitive edge in the green sector.Consequently, for SMEs, the impact of digital finance-induced spillover is more pronounced in the realm of high-end green innovation, as opposed to low-end green innovation.

Discussion
Climate change is currently a major concern worldwide, especially in the context of fostering green innovation within businesses.Building upon the transition from extensive to intensive economic development in emerging economies, this study explores the impact of financial reform services in the context of the digital economy on high-end green innovation within enterprises.Utilizing a unique perspective on knowledge spillovers, this theoretical research aims to unravel the 'black box' of production factors in information exchange on digital platforms.For countries with significant carbon emissions, this paper aims to alleviate the dilemma of low-end lock-in for green innovation within enterprises through digital finance and assist businesses in various countries in assuming their environmental governance responsibilities.However, existing research on digital finance metrics primarily focuses on the regional level.For instance, Syed et al (2021) employ the percentage of mobile money transactions to GDP and the number of ATMs per million people in certain South African countries as proxy variables for measuring digital finance.Al-Smadi (2023) measures the digital finance index using the number of commercial bank deposit accounts per 1000 adults in 12 Middle Eastern and North African countries.Many studies focusing on China widely adopt the Peking University Digital Financial Inclusion Index as a measure of digital finance (Rao et al 2022).This study establishes, for the first time, a micro-vocabulary at the enterprise level for measuring digital finance development.Through text analysis, key terms from the MD&A section of annual reports of listed companies are extracted and summarized as representatives of enterprise digital finance development, making a valuable attempt to fill the existing research gap.This not only aids in constructing a database for the development of digital finance in emerging economies, with China as a representative, but also provides empirical evidence for research on digital finance development in other economies.
At the same time, this study has certain limitations.Although existing research mainly relies on patent filings to measure the level of green innovation within enterprises, it's acknowledged that the number of patent filings may not fully reflect a company's green innovation capabilities (Gambardella et al 2008).Many of the patents filed by companies represent incremental innovations rather than breakthrough innovations, and companies may adopt defensive patent filing strategies to protect existing products and technologies.However, the number of green invention patent applications by companies to some extent reflects their emphasis on environmental responsibility and their willingness to innovate in green technologies.Therefore, considering these factors, this study chooses invention patents as the indicator to measure a company's green innovation capabilities.However, in future research, to overcome the limitations of measuring green invention patents, it may be worth considering measures such as green breakthrough innovations or utilizing methods such as patent citation frequency or knowledge depth to assess the quality of a company's green innovation.Additionally, this study relies on data from China, representing an emerging economy, and while the results are applicable within the context of China, it remains unknown whether the spillover effects of digital finance on green innovation within enterprises hold true in other economies.Future research could explore multinational data to enrich theoretical studies on factors influencing green innovation.

Conclusions and suggestions
The development of digital finance compared to traditional financial models exhibits significant knowledge spillovers and is an important pattern of digital economic development, playing a proactive role in fostering green innovation within enterprises.Following a series of empirical tests and path mechanism examinations, the following conclusions were drawn: Firstly, the development of digital finance within enterprises demonstrates evident knowledge spillover effects.Through the analysis of production functions incorporating knowledge spillover effects, it can be concluded that the development of digital finance within enterprises effectively promotes the overall development of green innovation.This effect is more pronounced in fostering high-end green innovation, thereby effectively alleviating the phenomenon of 'low-end lock-in' in green innovation among Chinese enterprises.
Secondly, empirical analysis results indicate that the development of digital finance within enterprises can effectively alleviate financing constraints and enhance internal control levels.Consequently, it improves the level of high-end green innovation by enhancing external resource acquisition and internal management capabilities.
Thirdly, the promotion effect of digital finance development on enterprise green innovation exhibits notable heterogeneity across different types of enterprises.Regarding corporate property rights, state-owned enterprises can promote green innovation output through digital finance development, with significant promotion effects observed in overall, high-end, and low-end green innovation.In terms of the region to which enterprises belong, digital finance development in the central and eastern regions can promote enterprise green innovation, while there is no evidence of promotion effects in the western region.Concerning enterprise scale, large-scale enterprises show significant promotion effects, whereas among small and medium-sized enterprise samples, only high-end green innovation exhibits promotion effects.The promotion effect of high-end green innovation in different types of enterprises is significantly superior to that of low-end green innovation.Therefore, the efficiency of knowledge spillover utilization brought about by digital finance development is higher in stateowned enterprises, central regions, and large enterprises, effectively mitigating the 'low-end lock-in' effect in enterprise green innovation.
Based on the research conclusions, the following recommendations are proposed: Firstly, the financial services reform departments of various economies need to provide policy support and experimentation space for digital financial services, and establish green knowledge exchange platforms.On one hand, governments should continue to promote digital infrastructure, cultivate digital finance professionals, and improve supporting industrial development.On the other hand, constructing digital financial knowledge service platforms, especially intra-enterprise green technology exchange activities, can be facilitated through relevant information platforms.Utilizing the information technology of digital finance development enables diversified technological iteration and upgrades.External knowledge exchange can lead to complementary knowledge, fundamentally breaking through technological barriers.
Secondly, governments of emerging economies need to guide enterprises towards green innovation and create a favorable environment for green technological upgrades.Enterprises need to promote stringent financial management systems, fully utilize the convenient financing services provided by digital finance development, dynamically adjust financing leverage to alleviate financing constraints, and utilize the information transmission of digital finance technology to enhance internal management levels, thereby promoting efficient enterprise operations.
Thirdly, China needs to implement policies based on enterprise heterogeneity and actively leverage the impact of digital finance on enterprise green innovation.Regarding corporate ownership, further promoting the green upgrading and transformation of state-owned enterprises, actively guiding enterprises to engage in highend green innovation, enhancing the green innovation awareness of state-owned enterprises, and encouraging non-state-owned enterprises to unleash green innovation vitality through digital finance are recommended.Concerning regions, the central and eastern regions should rely on mature financial markets and talent aggregation advantages to expand the green innovation effects of digital finance, while the western region should continuously improve digital infrastructure construction to narrow the digital divide between regions.In terms of enterprise scale, large enterprises should actively introduce digital finance financing models with their strong organizational operational efficiency to achieve high-end green output, while small and medium-sized enterprises should focus on green high-end innovation using digital finance development models to foster highquality green competitive advantages within enterprises.

Table 3 .
Empirical results of the main regression.

Table 4 .
Empirical results of PSM.

Table 7 .
Regression empirical results with the addition of control variables.

Table 8 .
Regression empirical results of the mechanism test.* and *** are significant at the level of 1%, 5%, and 10%, respectively.Robust t-statistics in parentheses.

Table 9 .
Regression empirical results on the heterogeneity of the properties of firms.

Table 10 .
Regression empirical results of regional heterogeneity of enterprises.

Table 11 .
Regression empirical results of firm size heterogeneity.