Impacts of Digital Transformation on the Quantity and Quality of Corporate Green Innovation: Evidence from China

Abstract

Green innovation is a key driver for reshaping the economic development model and attaining sustainable development. Prior research has primarily concentrated on the consequences of digital transformation on the quantity of corporate green innovation, while in-depth investigations into its effects on the quality of green innovation and the underlying mechanisms remain limited. Using data from Chinese A-share listed companies in Shanghai and Shenzhen between 2010 and 2023, this study empirically investigates the effects of digital transformation on both the quantity and quality of corporate green innovation. The results of this research show that, firstly, digital transformation exerts a substantial influence on both categories of green innovation. Second, the mechanism analysis reveals that, on the one hand, digital transformation promotes green innovation by alleviating information asymmetry, thereby effectively improving both its quantity and quality; on the other hand, it facilitates an increase in the quantity of green innovation through resource integration effects, but fails to enhance its quality via this channel. Third, the heterogeneity analysis shows that the positive impacts of digital transformation on the quantity and quality of green innovation are more pronounced in state-owned enterprises, firms located in eastern regions, and high-tech enterprises. This study enhances the research on how digital transformation can augment the quantity and elevate the quality of corporate green innovation, offering a valuable reference for advancing the sustainable development of organizations.

1. Introduction

Green innovation has evolved into an essential driving force for major economies globally to promote economic transformation and move towards sustainable development. Amid worsening global environmental pollution, accelerating the green transformation of firms is essential. (Bai et al., 2023; Jin et al., 2022) [1,2]. As a major engine of global economic growth, China is also facing severe challenges from increasing environmental pollution and excessive greenhouse gas emissions in the process of achieving rapid economic growth (Hu et al., 2023) [3]. To address these challenges, China is actively promoting a green development strategy (Jiang and Raza, 2023; Zhang et al., 2024) [4,5]. As the core of the green development strategy, green innovation is being recognized as an essential option for addressing global environmental issues. (Han et al., 2024; Wang and Sheng, 2024) [6,7]. Green innovation can effectively reduce the environmental risks of traditional technologies, reduce pollution emissions at the source, and boost energy efficiency (Bai et al., 2023) [1]. Nevertheless, the realization of green innovation encounters numerous challenges. Due to its high investment, high risk, and high uncertainty, enterprises often face cost pressure, technological integration difficulties, and limited financing channels when carrying out green innovation (Chen et al., 2022) [8]. These factors are intertwined, resulting in enterprises lacking sufficient internal motivation to pursue green innovation (Xu et al., 2024) [9]. Meanwhile, digital technology is developing rapidly, and digital transformation is steadily turning into a key driver of innovation and transformation among global enterprises (Tang et al., 2024) [10]; the advent of digital transformation has created fresh opportunities for enterprises to advance green innovation (Cao, 2023) [11]. With its high dependence and integration characteristics, digital transformation has shown unique advantages in transforming production processes and improving resource utilization efficiency (Liu et al., 2023) [12]. In addition, digital technology can effectively integrate technology, data, and knowledge resources, which not only transforms the traditional innovation model of enterprises and the combination of innovation elements, but also plays a crucial part in alleviating the financial pressure of companies throughout the procedure of sustainable innovation, breaking the barriers of information asymmetry and helping firms to overcome technical difficulties, thereby facilitating conducive environments for firm green innovation (Luo et al., 2023) [13].

At present, studies examining how digital transformation affects corporate green innovation can be extensively separated into three viewpoints. Firstly, digital transformation positively promotes corporate green innovation (Shen and Tan, 2022; Liu et al., 2023) [14,15]. Mubarak et al. (2021) [16] discovered that the implementation of fourth industrial revolution technologies may boost the green innovation capacities of organizations. Song et al. (2022) [17] demonstrated that digital transformation greatly enhances the green innovation of extremely polluting industries. Secondly, in the course of digital transformation, firms may reallocate internal resources, which increases cost pressures and, ultimately, reduces their willingness to engage in green innovation (Liu et al., 2024; Yang et al., 2024; Du and Cao, 2023) [18,19,20]. Thirdly, other studies have suggested that they may follow a nonlinear inverted U-shaped pattern, with the relationship initially increasing, and then decreasing (Ning et al., 2024) [21].

The researchers examined the mechanisms between them from two different perspectives: external and internal. At the external environment level, existing research shows that, when companies face stronger external regulatory pressure, greater government innovation subsidies, and higher media attention, the impact becomes increasingly significant. (Zhang et al., 2025; Guo et al., 2023) [22,23]. At the internal factor level, prior studies have primarily focused on how digital transformation boosts green innovation through diminishing communication obstacles, mitigating agency conflicts, and strengthening internal control (Zhang et al., 2021; Li et al., 2023; Duan et al., 2024) [24,25,26].

Relative to extant studies, the incremental benefit provided by our study is evident in each of the following three dimensions: Firstly, in terms of research data, prior studies have predominantly measured corporate digital transformation by the ratio of hardware investments in digital technologies to total assets (Song et al., 2022) [17]. However, this indicator neglects both the outputs and actual effectiveness of digital transformation, as well as its multidimensional nature. To address this limitation, this study employs Python (3.8.8)-based text mining to examine the yearly documents of companies that are listed on A-shares spanning from 2010 to 2023, establishing a metric for digital transformation by quantifying the frequency of terms associated with “digital transformation” in the reports. Secondly, in terms of research perspective, while earlier studies have primarily concentrated on the quantity of green innovation, they have overlooked its influence on the quality of green innovation (Li et al., 2024) [27]. The quality of green innovation plays an essential role in facilitating a firm’s green transformation and ensuring long-term sustainable development. We perform a comprehensive analysis from both the “innovation quantity” and “innovation quality” dimensions, examining the effects of digital transformation on each. Thirdly, in terms of research content, this article enriches the influence mechanism between them. Whereas existing research has mostly centered on pathways such as environmental regulation, financial performance, and internal control, providing insufficient insights into the underlying mechanisms (Yang et al., 2025; Liu et al., 2023) [28,29], this study highlights two key channels: information interaction and resource integration.

2. Theoretical Analysis and Research Hypothesis

As key actors in promoting sustainable development, enterprises can benefit from green innovation, which is essential for their ecological transition and sustainable advancement. Digital transformation enables enterprises to efficiently integrate and utilize knowledge and information, drive the restructuring of business processes and production methods, optimize the efficient allocation of production factors, and further stimulate their green innovation (Sun et al., 2024) [30]. First, the extensive adoption of online technologies like big data, blockchain, and the Internet within enterprises has facilitated easier access to advanced patents. Digital technologies can also transcend the constraints of time and space, facilitating the rapid dissemination of knowledge and information within enterprises’ green innovation activities (Li et al., 2022) [31]. Second, digital technologies can help optimize corporate organizational structures and business processes, thereby improving resource utilization efficiency (Stoenoiu and Jäntschi, 2024) [32] and reducing operating and sales costs (Gil-Alana et al., 2020) [33], allowing enterprises to dedicate greater resources to green innovation initiatives. Based upon the preceding discussion, we put forward the following hypothesis:

Hypothesis 1:

Digital transformation can promote both the quantitative expansion and qualitative improvement of corporate green innovation.

2.1. Resource Integration Effect

The natural resource-driven concept asserts that internal organizational variables, including a firm’s technological capabilities, form the foundation for company green innovation. Consequently, digital transformation can influence green innovation by improving the distribution and exploitation of creative resources. (Wang et al., 2021) [34]. Firstly, from the standpoint of resource allocation, digital technology can promptly identify enterprise needs and accurately match supply and demand. Digital technology precisely allocates enterprise resources, such as capital and raw materials, to green projects with significant economic benefits, rational resource utilization, and coordinated environmental and economic development. This allows for a more rational allocation of existing resources, effectively promoting green innovation activities within enterprises. Secondly, considering the standpoint of resource integration, green innovation encompasses knowledge and information from diverse areas, including manufacturing, pollutant mitigation, and consumption of energy reduction (Luo et al., 2025) [35]. The green innovation process requires integrating knowledge from diverse organizational domains to accurately grasp key technologies for green technological innovation. Digital transformation can effectively promote collaborative and open innovation (Zhang et al., 2019) [36], broaden the range of innovation resource deployment, promote collaborative innovation among firms, and augment their capacity for individual technological advancement utilizing currently available technology. Third, from a resource-sharing perspective, digital transformation can facilitate the integrated sharing of R & D resources, achieve the seamless integration of R & D personnel and assets, and enhance the exchange and reconfiguration of knowledge elements across different technological domains. (Xiong et al., 2025) [37]. Based upon the preceding discussion, we put forward the following hypothesis:

Hypothesis 2:

Digital transformation fosters corporate green innovation by improving resource allocation efficiency.

2.2. Information Interaction Effect

Existing studies have shown that green innovation not only requires enterprises to effectively integrate resources, but also places higher demands on enterprises’ information sharing capabilities (Zhao et al., 2020) [38]. Information asymmetry can lead to potential conflicts of interest between shareholders and senior executives, negatively impacting a company’s green innovation activities (Hoang et al., 2020) [39]. From an internal perspective, digital transformation facilitates cross-departmental information exchange, making management and operations more transparent and information more accessible. This not only simplifies business processes and strengthens the integration and sharing of internal information, but also enhances a company’s innovation capabilities (Lin et al., 2024) [40]. Furthermore, the digitization of management processes helps reduce management’s self-interest, thereby reducing agency conflicts between management and shareholders. From an external perspective, digital technology can better enable market mechanisms (Adra and Barbopoulos, 2019) [41], break down traditional information barriers, enhance information oversight, strengthen external regulatory mechanisms, and boost investor confidence and motivation. Furthermore, digital transformation can promote real-time information sharing, improve supply chain collaboration, and enhance a company’s ability to obtain external financial resources (Qi and Xiao, 2020) [42]. Based upon the preceding discussion, we put forward the following hypothesis:

Hypothesis 3:

Digital transformation lowers information asymmetry of firms, enhances their information exchange capabilities, and promotes engagement in green innovation.

In summary, existing research has primarily analyzed green innovation from the perspective of quantity (Li et al., 2024) [27], while this paper thoroughly examines the significance of digital transformation on its quantity and quality. Meanwhile, current studies have primarily analyzed mechanisms from the perspectives of constraints on financing and internal control (Yang et al., 2025; Liu et al., 2023) [28,29]. From the perspectives of information interaction and resource integration, this paper explores the internal mechanism between the two and expands the theoretical framework and path. Therefore, we examine the effects and underlying mechanisms of them by utilizing data from Chinese A-share listed companies and employing methods of text mining to create indicators related to digital transformation. A multivariate fixed-effects model is utilized for empirical analysis.

3. Materials and Methods

This study initially collected data from Chinese A-share listed companies on the Shanghai and Shenzhen Stock Exchanges for the period 2010–2023. To ensure data quality and validity, this paper removed ST and ST* firms, firms in the financial and insurance sectors, as well as samples exhibiting significant missing signs or anomalous values in the control variables. Additionally, we also winsorized all continuous variables at the first and 99th percentiles to mitigate the impact of extreme values. The total dataset has 43,616 firm–year observations. Financial data for A-share listed companies was obtained from the China Stock Market & Accounting Research (CSMAR) database; meanwhile, information on corporate green invention patents and their citations was obtained from the Green Patent Research Database (GPRD). The corporate digital transformation was manually collected from the annual reports of the listed companies throughout the sample period.

The dependent variable is measured using two indicators: the quantity and the quality of corporate green innovation. First, for the quantity of corporate green innovation (lnGreenInnovation), since green patents represent the highest level of technological content in corporate green knowledge, following the existing literature (Messeni et al., 2011; Zhang et al., 2024) [43,44], we quantify green innovation intensity by the count of green invention applications for patents and utilize the natural logarithm of this value plus one to mitigate data skewness and accommodate instances with zero observations. Second, the quality of corporate green innovation (lnCite), as patent citation frequency is an important indicator of the technological impact of a patent, following previous studies (Xu et al., 2024) [9], we utilize the citation count of green invention patents as an indicator of green innovation quality and apply the natural logarithm of this count plus one to normalize the data and mitigate the impact of outliers.

The core explanatory variable of this paper is digital transformation (lnDigital). Drawing on the studies by Tang et al. (2023), Wu et al. (2021), and Yuan et al. (2024) [10,45,46], we assess the extent of the core explanatory variable by quantifying the total frequency of digital transformation-related terms in the yearly documents of publicly traded companies. Specifically, we first constructed a dictionary of enterprise digitalization terms. Combining relevant academic research and economic policy text documents, 173 digitalization-related terms were selected from the four areas of technology—big data, cloud computing, artificial intelligence, and blockchain—to form an enterprise digitalization terminology dictionary. Secondly, text mining was performed through Python software to systematically arrange the yearly statements of A-share publicly traded enterprises from 2010 to 2023. With the help of the PDFplumber library, all PDF files were batch converted into txt format, and all text content was extracted to provide basic data for subsequent feature word screening. Next, given the complexity of annual report content and the tendency for companies to present their business operations, development plans, and related information in the “Management Discussion and Analysis” (MD & A) section, we concentrated on analyzing the text in this section. After splitting the MD&A section, we divided the sentences into individual word tokens, and for each occurrence of the keyword “digital transformation,” anchored it and examined a fixed-length window of words to its left. If any negation terms (e.g., “not,” “no,” “none”), or words referring to entities other than the focal firm (e.g., the company’s shareholders, customers, suppliers, or executives), appeared within this window, we marked the keyword for deletion; otherwise, we retained it. In this way, expressions containing negation words preceding the keyword, as well as instances of “digital transformation” that did not refer to the focal firm itself, were removed. Next, the Jieba library was used to extract keywords from the specified text file and count the frequency of occurrence of these keywords. Finally, the number of keyword occurrences in the four dimensions of enterprise digital transformation was summarized to obtain the aggregate word frequency of enterprise digital transformation. To eliminate the influence of outliers, we logarithmized the total word frequency to obtain quantitative indicators that characterize enterprise digital transformation.

To control other influences and more precisely assess the effect of digital transformation on corporate green innovation, drawing on the research of Chen et al. (2022), Cao et al. (2022), and Wang et al.(2021) [8,11,34], the following variables for control were chosen based on enterprise characteristics, financial condition, and corporate governance capacity: enterprise size (Size), company establishment year (FirmAge), asset–liability ratio (Lev), profit from assets (ROA), cash flow to asset ratio (Cash-flow), net profit growth rate (NetProfitGrowth), management shareholding ratio (Mshare), the share of independent directors (Indep), and Tobin’s Q ratio (TobinQ). The precise definitions of the primary variables are presented in

Table 1. Definitions of main variables.

Variable	Variable Symbols	Definition
Green innovation	lnGreenInnovation	The natural log of (the number of green invention applications plus one)
	lnCite	The natural log of (the citation counts of green invention patents plus one)
Digital transformation	LnDigital	The natural log of the sum frequency of phrases of digital transformation
Company size	Size	Natural logarithm of total assets per year
Years of enterprise	FirmAge	Ln (current year minus company establishment year plus 1)
Asset–liability ratio	Lev	Year-end total liabilities divided by year-end total assets
Net return on assets	ROA	Net profit/average balance of total assets
Cash flow ratio	Cashflow	Net operational cash flow divided by total value of assets
Net profit growth rate	NetProfitGrowth	(Net profit this year/net profit last year) minus 1
Management shareholding ratio	Mshare	Executive-held shares split by total shares
the share of independent directors	Indep	Independent Directors/Number of Directors
Tobin’s Q	TobinQ	(Market value plus Bank worth of liabilities) divided by total assets

.

This study employs regression analysis using a multi-dimensional fixed-effects model. The first explained variable is given by the quantity of green innovations of enterprises (lnGreenInnovation), and its model is shown in Formula (1), as follows:

$$lnGreenInnovatio{n}_{ijst}={\beta }_0+{\beta }_{1}ln{\mathrm{Digital}}_{ijst}+{\gamma }^{\prime }{X}_{ijst}+{\lambda }_i+{\theta }_{jt}+{\mu }_{st}+{\epsilon }_{ijst}$$

(1)

The second explained variable is the enterprise green innovation quality (lnCite), and its model is shown in Formula (2), as follows:

$$lnCit{e}_{ijst}={\beta }_0+{\beta }_{1}ln{\mathrm{Digital}}_{ijst}+{\gamma }^{\prime }{X}_{ijst}+{\lambda }_i+{\theta }_{jt}+{\mu }_{st}+{\epsilon }_{ijst}$$

(2)

In the above two models, the subscripts i, j, s, t represent the company, industry, province, and time, respectively. ${\beta }_0$ is a constant term; $ln{\mathrm{Digital}}_{ijst}$ is the principal explanatory variable of this paper, representing the extent of digital transformation of enterprises in the s province and j industry in t year; $X_{ijst}$ signifies a set of control variables; ${\lambda }_i$ represents the firm fixed effect. To mitigate the influence of time-varying unobservable characteristics at both the industry and provincial levels, we also incorporated industry–year fixed effects ${\theta }_{jt}$ and province–year fixed effects ${\mu }_{st}$; ${\epsilon }_{ijst}$ represents the random disturbance term, for which the clustered robust standard error at the industry–year level is used. All regression results were completed using STATA17 software.

We used a multivariate fixed-effects model for analysis. Compared to ordinary least squares (OLS) regression, the multivariate fixed-effects model can control firm-level heterogeneity while simultaneously eliminating the confounding of time-varying unobservable factors across industries and regions. Corporate innovation initiatives are influenced by numerous factors, including industry cycles and regional policy discrepancies. Traditional OLS regression alone would be susceptible to omitted variable bias. The multivariate fixed-effects model effectively controls these potential confounding factors and eliminates the influence of time-varying firm-level idiosyncrasies, such as historical R & D accumulation. Industry–year fixed effects control for common shocks faced by firms in different industries across different years (such as technology cycles and industry policy adjustments). Province–year fixed effects control for regional and year-to-year differences in the macroenvironment (such as local government green development policies and the intensity of regional environmental regulation). By employing these multidimensional controls, the model minimizes the confounding of unobservable variables on the estimation results.

This paper examines the validity of the regression coefficients. (When the coefficient ${\beta }_1$ results are significant, it demonstrates that digital transformation strongly influences corporate green innovation, thus verifying Hypothesis 1. Secondly, to evaluate the overall explanatory strength of the regression model, this study presents the adjusted R-squared value (Adjusted R²). Furthermore, in the robustness check section, this study further validates the multivariate fixed-effects results by substituting the dependent variables, replacing the key explanatory variables, modifying the model estimation method, and applying instrumental variable analysis to strengthen the reliability of the findings.

4. Results

4.1. Descriptive Statistics

In

Table 2. Descriptive statistics of main variables.

Variable	Observations	Mean	Std. Dev.	Min	Max
GreenInnovation	43,636	0.240	0.640	0	6.328
Cite	43,636	0.539	0.986	0	4.174
Digital	43,636	1.480	1.422	0	5.357
Size	43,636	22.170	1.300	19.590	26.440
FirmAge	43,636	0.412	0.208	0.0274	0.925
Lev	43,636	0.041	0.067	−0.375	0.255
ROA	43,636	0.046	0.069	−0.226	0.267
Cashflow	43,636	−0.378	3.626	−37.030	14.890
NetProfitGrowth	43,636	0.377	0.053	0.286	0.600
Mshare	43,636	2.008	1.316	0.795	15.610
Indep	43,636	2.921	0.350	1.099	3.638
TobinQ	43,636	0.150	0.204	0	0.709

, the mean number of green invention patents (GreenInnovation) is 2.183, with a standard deviation of 9.329. The average citation count of green invention patents (Cite) is 3.559, with a standard deviation of 41.363. It is not difficult to see that, in terms of both quantity and quality, the general degree of green innovation across firms is quite low, with significant disparity between enterprises. The average digital transformation (Digital) score is 13.93, with a standard deviation of 28.905. This indicates that digital transformation appeared in corporate annual reports only 13.93 times between 2010 and 2023, with significant variation among enterprises.

4.2. Baseline Regression Results

Columns (1) to (3) in

Table 3. Benchmark regression results.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	lnGreenInnovation	lnGreenInnovation	lnGreenInnovation	lnCite	lnCite	lnCite
LnDigital	0.026 ***	0.022 ***	0.018 ***	0.032 ***	0.016 ***	0.015 ***
	(0.003)	(0.004)	(0.004)	(0.006)	(0.006)	(0.005)
Size		0.039 ***	0.045 ***		0.138 ***	0.142 ***
		(0.006)	(0.006)		(0.013)	(0.011)
FirmAge		0.008	−0.090 **		0.292 ***	0.111 *
		(0.040)	(0.042)		(0.067)	(0.060)
Lev		−0.011	−0.025		0.100 ***	0.056 *
		(0.020)	(0.019)		(0.038)	(0.032)
ROA		0.049	0.058		−0.257 ***	−0.280 ***
		(0.046)	(0.048)		(0.078)	(0.077)
Cashflow		−0.038	−0.027		0.164 ***	0.204 ***
		(0.033)	(0.034)		(0.059)	(0.049)
NetProfitGrowth		0.000	0.000		0.000	−0.000
		(0.001)	(0.001)		(0.001)	(0.001)
Mshare		0.021	0.048 *		−0.120 ***	−0.097 *
		(0.025)	(0.027)		(0.044)	(0.054)
Indep		0.064	0.052		−0.088	−0.077
		(0.062)	(0.063)		(0.093)	(0.091)
TobinQ		0.007 ***	0.008 ***		0.009 **	0.012 ***
		(0.002)	(0.002)		(0.004)	(0.004)
Constant	0.202***	−0.722 ***	−0.561 ***	0.497 ***	−3.375 ***	−2.941 ***
	(0.006)	(0.170)	(0.170)	(0.013)	(0.369)	(0.307)
Individual fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	No	Yes	Yes	No
Province fixed effects	Yes	Yes	No	Yes	Yes	No
Industry–year fixed effects	No	No	Yes	No	No	Yes
Province–year fixed effects	No	No	Yes	No	No	Yes
Adjusted R-squared	0.643	0.654	0.657	0.725	0.739	0.762
Observations	43,298	38,583	38,562	43,380	38,626	38,605

Note: Significance at the 1%, 5%, and 10% thresholds is marked by ***, **, and *, respectively; the values in brackets are robust standard errors.

display the regression-based outcomes. Column (1) displays the baseline specification, which includes only digital transformation along with firm, province, and year fixed effects. Column (2) extends the baseline model by adding control variables, including firm size (Size), firm age (FirmAge), leverage (Lev), return on assets (ROA), cash flow ratio (Cashflow), net profit growth rate (NetProfitGrowth), management shareholding ratio (Mshare), proportion of independent directors (Indep), and Tobin’s Q (TobinQ). Industry–year and province–year fixed effects are incorporated in column (3). Columns (4) through (6) display the regression outcomes of the quality of corporate green innovation, with the column order and variable settings consistent with those in Columns (1) to (3). The outcomes displayed in columns (1) and (4) of Table 3 demonstrate that, without adding any control variables, digital transformation exerts a substantial influence on both categories of green innovation. Columns (2) and (5) of Table 2 present the regression outcomes after the inclusion of control variables. The coefficients of the key explanatory variables remain significantly positive. Finally, considering that certain characteristics of enterprises in different industries and provinces may change over time, this paper further introduces industry–year fixed effects and province–year fixed effects, to more finely control the impacts of unobservable factors at the industry and provincial levels that change over time. In columns (3) and (6) of Table 3, after including the aforementioned control variables, industry–year fixed effects and province–year fixed effects, the coefficients of the regression are 0.018 and 0.015, respectively. Enhancing digital transformation by 1% results in a 0.018% growth in the quantity of corporate green innovations and a 0.015% improvement in quality. The outcomes substantiate Hypothesis 1 empirically. The results demonstrate that digital transformation exerts a substantial influence on both categories of green innovation. Significantly, the effect of it is more pronounced in augmenting the total quantity of green innovation activities than in enhancing the quality.

4.3. Robustness Tests

To further validate the accuracy of the model findings, this study utilizes the following approaches of changing the explained variable, one-period lagged processing, changing the model setting, instrumental variable method, Heckman two-step method, and so on to conduct the robustness test.

4.3.1. Replacing the Explained Variable

To confirm the accuracy of the model results, we replaced the explained variables with “the number of green patent applications (Green_Total)” and the number of green patents cited by other companies after excluding the number of self-citations (lnCite_others). The findings after changing the explained variables are shown in columns (1) and (2) of

Table 4. Regression results of replacing the explained variable and lagged one period.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	Green_Total	lnCite_Others	lnGreenInnovation	lnCite	lnGreenInnovation	lnCite
LnDigital	0.018 ***	0.017 ***
	(0.004)	(0.004)
Dig			1.805 ***	2.544 ***
			(0.541)	(0.667)
L.LnDigital					0.019 ***	0.027 ***
					(0.004)	(0.005)
Control	Yes	Yes	Yes	Yes	Yes	Yes
Individual fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Industry–year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Province–year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Adjusted R-squared	0.670	0.750	0.663	0.808	0.668	0.773
Observations	38,562	38,605	32,985	33,005	35,920	35,938

Note: Significance at the 1% threshold is marked by ***; the values in brackets are robust standard errors. The chosen variables for control are the same as those in column (6) of Table 3.

. Therefore, these results strengthen the validity of the baseline regression results.

4.3.2. Replacing the Explained Variable

Since the total word count in annual reports differs across companies, measuring digital transformation using the aggregate frequency of associated terms may be influenced according to the lengths of their reports. To avoid estimation bias, we replaced the principal explanatory variable with the “ratio of the total word frequency of digital transformation to the total word frequency of the annual report” (Dig) to more accurately represent the true level of digital transformation of the company. The regression results using an alternative core explanatory variable are presented in columns (3) and (4) of Table 4. These results indicate that digital transformation continues to have a significantly positive effect on corporate green innovation, further confirming the robustness of the findings.

4.3.3. Explanatory Variables Lagged One Period

Since there may be a time lag in the output procedure of green patents from invention, application, and approval to citation, and, simultaneously, digital transformation is a gradual process, the influence it has on the overall number of green invention patents and patent citations may exhibit a time lag. This paper delayed the company’s digital transformation variables by one period for robust testing. The regression outcomes, presented in columns (5) and (6) of Table 4, demonstrate that LnDigital exerts a statistically substantial beneficial effect on green patents. This shows that, after considering possible time lag effects, enterprise digital transformation persistently enhances the green innovation of companies. Additionally, the regression coefficients for the first lag period exceeded those of the benchmark regression, showing that the promoting effect may exhibit an upward trend over time. In essence, as enterprise digital transformation deepens progressively, its influence on fostering green innovation development also intensifies gradually.

4.3.4. Change Model Settings

Given that the quantities of green innovation patents and green patent citations are both count data, and there are many zero values, Poisson pseudo-maximum likelihood regression (PPML) is a regression method used to estimate count data models. It assumes that the dependent variable follows Poisson distribution and solves the model parameters through maximum likelihood estimation. PPML is not only applicable to count data, but can also be applied to any non-negative dependent variable, especially when there are many zero values. Based on this, this paper uses Poisson pseudo-maximum likelihood regression for robustness testing and replaces the explained variables with the number of green invention patents and the citation frequency of green patents. The regression results are shown in columns (1) and (2) of

Table 5. Regression of replacement model.

Variable	(1)	(2)	(3)	(4)
	Number of Patents	Patent Citations	lnGreenInnovation	lnCite
LnDigital	0.237 ***	0.146 ***	0.126 ***	0.046 ***
	(0.030)	(0.032)	(0.030)	(0.013)
Control	Yes	Yes	Yes	Yes
Individual fixed effects	Yes	Yes	Yes	Yes
Industry–year fixed effects	Yes	Yes	Yes	Yes
Province–year fixed effects	Yes	Yes	Yes	Yes
Adjusted R-squared	0.1325	0.1995	0.7926	0.9198
Observations	38,979	39,013	17,356	14,677

Note: Significance at the 1% threshold is marked by ***; the values in brackets are robust standard errors. The chosen variables for control are the same as those in column (6) of Table 3. Both the Tobit model and Poisson pseudo-maximum likelihood regression (PPML) report the pseudo-coefficient of determination.

. Digital transformation exerts a substantial influence on both categories of green innovation.

Since the quantity and quality of corporate green patent applications are left-truncated at 0, to avoid possible bias, this paper further adopts the Tobit model for robustness testing. The regression outcomes, displayed in columns (3) and (4) of Table 5, show that the coefficients remain strongly positive, further confirming the robustness of this study’s conclusions.

4.3.5. Instrumental Variables

Considering the possibility of a bidirectional causal relationship between them, companies with abundant innovation resources and high-quality innovation platforms are more likely to actively promote digital transformation, due to their strong R&D capabilities and sufficient financial support. To solve this endogeneity problem, this study adopts the methodology of Huang et al. (2019) [47] and uses the 1984 city-level ratio of fixed-line telephones per 100 residents as an instrumental variable for the explanatory variables. From a correlation perspective, 1984 was the early days of China’s transition. As a long-distance communications infrastructure, the regional distribution of fixed-line telephones directly determined the coverage density and penetration of early regional communications networks. On the one hand, the early installation and layout of fixed-line telephone lines provided the network foundation for digital technology (Yuan et al., 2021) [47]. From another perspective, based on usage habits, businesses in an environment with widespread fixed-line telephones are more receptive to digital networks, showing that the selected instrumental variables meet the principle of correlation. From an exogenous perspective, the layout of China’s fixed-line telephones in 1984 was primarily determined by national communications infrastructure planning. Most A-share listed companies were either not yet established or were in their early stages of development. Furthermore, the core function of fixed-line telephones in 1984 was to meet basic voice communication needs. Fixed-line telephones themselves did not involve green technology R&D, environmental protection investment, or other aspects directly related to green innovation. Therefore, it was difficult to indirectly influence a firm’s current green innovation decisions through policy support or resource allocation and, thus, did not directly affect their current green innovation performance. Therefore, the exogeneity requirement was met. However, since the 1984 city-level count of fixed-line telephones per 100 households is cross-sectional and remains constant over time, directly applying it to panel data is likely to lead to weak instrumental variable bias (Bound et al., 1995) [48]. Moreover, the number of fixed-line telephones, as a static historical indicator, cannot directly explain the dynamic diffusion process of Internet development (Czernich et al., 2011) [49]. To overcome this limitation, this paper introduces a dynamic indicator—the total count of Internet users across the nation in the preceding year—and creates an interaction term between the two to serve as a tool variable.

Simultaneously, we performed a weak instrument test. As shown in Table 5, the Cragg–Donald Wald F statistics surpass the Stock–Yogo 10% threshold value, and the first-stage F statistics are all above 10, indicating that weak instrument issues are not present. In addition, the test for under-identification indicates that the p-value of the Kleibergen–Paap rk LM statistic is below 0.01, which means that the instrumental variable is identifiable.

Table 6. Instrumental variable regression results.

Variable	(1)	(2)	(1)	(2)
	LnDigital	lnGreenInnovation	LnDigital	lnCite
LnDigital		0.283 ***		0.626 ***
		(0.066)		(0.104)
IV	0.092 ***		0.092 ***
	(0.010)		(0.010)
Control	Yes	Yes	Yes	Yes
Individual fixed effects	Yes	Yes	Yes	Yes
Industry–year fixed effects	Yes	Yes	Yes	Yes
Province–year fixed effects	Yes	Yes	Yes	Yes
Observations	33,505	33,505	33,539	33,539
F	82.623	104.777	82.900	400.709
CD Wald F	71.711		71.955
KP rk LM	81.475		81.759
	(0.000)		(0.000)

Note: Significance at the 1% thresholds is marked by ***; the values in brackets are robust standard errors. The chosen variables for control are the same as those in column (6) of Table 3. The values in brackets in the KP rk LM results are P values.

displays the first-stage model findings in columns (1) and (3), while columns (2) and (4) show the second-stage outcomes. The coefficients for digital transformation become significantly positive, demonstrating that, even after mitigating endogeneity via the instrumental variable approach, the coefficients of the core explanatory variables are still significant.

4.3.6. Heckman Two-Step Method

Enterprises in different industries show different enthusiasm and conditions when applying for green patents. For example, since heavily polluting enterprises face stricter environmental regulations and higher social pressure, to meet environmental protection requirements, enhance corporate image, and obtain policy benefits, they are often more motivated to implement sustainable innovation and actively apply for green patents (Liu et al., 2020) [50]. In contrast, non-heavily polluting enterprises may attach less importance to green innovation and are relatively less motivated to apply for green patents.

This research uses the Heckman two-step method to mitigate potential sample selection bias and verify result correctness by examining endogeneity resulting from sample selection. In the initial phase, the dependent variable signifies the presence of green invention patents or green patent citations for a firm, coded as 1 if present and 0 if absent. All control variables in the baseline regression are added to construct the Probit model for regression analysis. Then the two inverse Mills ratios (IMR1 and IMR2) obtained in the first stage are added to two models as control variables for the second stage regression. The outcomes in

Table 7. Heckman two-step regression results.

Variable	(1)	(2)
	lnGreenInnovation	lnCite
LnDigital	0.019 ***	0.012 ***
	(0.004)	(0.004)
IMR1	2.274 ***
	(0.307)
IMR2		2.569 ***
		(0.140)
Control	Yes	Yes
Individual fixed effects	Yes	Yes
Industry–year fixed effects	Yes	Yes
Province–year fixed effects	Yes	Yes
Adjusted R-squared	0.659	0.815
Observations	38,496	33,729

Note: Significance at the 1% thresholds is marked by ***; the values in brackets are robust standard errors. The chosen variables for control are the same as those in column (6) of Table 3.

suggest the estimated coefficients of the primary explanatory variables stay strongly positive, consistent with the baseline regression outcomes. It signifies that the primary conclusions of this study persistently hold true, even when considering bias in the selection of samples.

4.4. Mechanism Analysis

Based on the theoretical analysis, there may be two mechanisms of information interaction and resource integration between them. This study uses the subsequent regression model to evaluate Hypotheses 2 and 3, as follows:

$$Y_{ijst}={\beta }_0+{\beta }_{1}ln{\mathrm{Digital}}_{ijst}+{\gamma }^{\prime }{X}_{ijst}+{\lambda }_i+{\theta }_{jt}+{\mu }_{st}+{\epsilon }_{ijst}$$

(3)

$${\mathrm{Menchanism}}_{ijst}={\beta }_0+{\beta }_{1}ln{\mathrm{Digital}}_{ijst}+{\gamma }^{\prime }{X}_{ijst}+{\lambda }_i+{\theta }_{jt}+{\mu }_{st}+{\epsilon }_{ijst}$$

(4)

$$Y_{ijst}={\beta }_0+{\beta }_{1}ln{\mathrm{Digital}}_{ijst}+{\beta }_2{\mathrm{Menchanism}}_{ijst}+{\gamma }^{\prime }{X}_{ijst}+{\lambda }_i+{\theta }_{jt}+{\mu }_{st}+{\epsilon }_{ijst}$$

(5)

Firstly, Equation (3) verifies the baseline regression. Secondly, Equation (4) examines the effect of digital transformation on the mechanism variables. Finally, Equation (5) verifies the mechanism of information interaction and resource integration between the two, thus verifying Hypotheses 2 and 3. The mechanism variables were selected as follows: First, information interaction (ASY): Following the approach of Song et al. (2021) [51], we first developed three stock liquidity measures: the liquidity ratio (LR), the illiquidity ratio (ILL), and the return reversal indicator (GAM). We then applied principal component analysis to these original indicators (LR, ILL, and GAM) to derive an information asymmetry measure (ASY). A higher value of this measure indicates a higher degree of information asymmetry. Second, resource integration (TFP_OLS): With reference to the research conducted by (Xiong et al., 2025) [37], we used the Cobb–Douglas production function and performed an OLS regression analysis on the linearized production function to obtain an estimated value of TFP_OLS.

4.4.1. Resource Integration Effect

This investigation employs total factor productivity (TFP) to assess the consequences of resource integration. The regression outcomes are displayed in

Table 8. Regression results of resource integration mechanism.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	lnGreenInnovation	TFP_OLS	lnGreenInnovation	lnCite	TFP_OLS	lnCite
LnDigital	0.0183 ***	0.0184 ***	0.0178 ***	0.0147 ***	0.0184 ***	0.0152 ***
	(0.0035)	(0.0034)	(0.0036)	(0.0046)	(0.0034)	(0.0048)
TFP_OLS			0.0120 *			−0.0068
			(0.0071)			(0.0104)
Control	Yes	Yes	Yes	Yes	Yes	Yes
Individual fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Industry–year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Province–year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Adjusted R-squared	0.657	0.946	0.660	0.762	0.946	0.765
Observations	38,562	36,975	36,966	38,583	36,975	36,972

Note: Significance at the 1% and 10% thresholds is marked by *** and *, respectively; the values in brackets are robust standard errors. The chosen variables for control are the same as those in column (6) of Table 3.

. In columns (2) and (5), the coefficients for digital transformation are highly positive at the 1% level, demonstrating that digitalization markedly improves enterprises’ resource integration efficiency. Column (3) indicates that digital transformation exerts a substantial influence on both categories of green innovation, however diminished in magnitude, meaning that digitalization fosters corporate green innovation primarily through the mechanism of resource integration. The outcomes shown in column (6) demonstrate that the coefficient of total productivity is insignificant, implying that the resource integration effect does not significantly influence the enhancement of corporate innovation quality through digital transformation. The possible reason is that the process of resource integration mainly solves the problem of “resource misallocation” and “process redundancy,” such as shortening the start-up time of R&D projects through data sharing, quickly replicating the existing technology path, and thus increasing the number of innovation outputs to meet the policy and market demand (Song et al., 2025) [52]. However, resource integration can only provide basic support for the breakthrough technology R & D (such as disruptive green materials and zero-carbon production processes) required for high-quality innovation, but cannot directly promote the transformation of the technological paradigm. Radical technological innovation requires long-term accumulation and high-risk trial and error, and the “efficiency orientation” of resource integration struggles to align with its long-term research requirements. Therefore, digital transformation may increase the quantity of green innovation exclusively through resource integration, but it is difficult to achieve a leap in green innovation quality.

4.4.2. Information Interaction Effect

This paper tests the above possible mechanisms based on the information asymmetrical index ASY constructed by text analysis. The results in columns (2) and (5) of

Table 9. Regression results of information interaction mechanism.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	lnGreenInnovation	ASY	lnGreenInnovation	lnCite	ASY	lnCite
LnDigital	0.0183 ***	−0.0054 ***	0.0179 ***	0.0147 ***	−0.0054 ***	0.0144 ***
	(0.0035)	(0.0020)	(0.0035)	(0.0046)	(0.0020)	(0.0046)
ASY			−0.0752 ***			−0.0472 **
			(0.0141)			(0.0191)
Control	Yes	Yes	Yes	Yes	Yes	Yes
Individual fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Industry–year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Province–year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Adjusted R-squared	0.657	0.794	0.658	0.762	0.794	0.762
Observations	38,562	38,577	38,553	38,583	38,577	38,574

Note: Significance at the 1%and 5% thresholds is marked by *** and **, respectively; the values in brackets are robust standard errors. The chosen variables for control are the same as those in column (6) of Table 3.

demonstrate that demonstrating that digital transformation can effectively mitigate the information asymmetrical issues encountered by enterprises. The coefficients for digital transformation in columns (3) and (6) are notably positive, with a diminished absolute value. That finding corresponds with the prior mechanism analysis, implying that digital transformation facilitates firms in assimilating resources by improving their information exchange capacities, thereby optimizing their innovative technology resources and motivating them to participate in sustainable innovation efforts, thereby significantly improving green innovation.

4.5. Heterogeneity Analysis

4.5.1. Heterogeneity of Property Rights

In contrast to other developed nations, Chinese firms are categorized into government and private entities based on differing ownership rights. There are great differences in the social responsibilities, resources, and policy support they bear, which subsequently influences their perspectives on digital transformation as well as their capabilities and motivations for green innovation. Building on this, this study conducts heterogeneity tests by grouping firms according to property rights. The results are presented in columns (1) to (4) of

Table 10. Regression results of property rights heterogeneity.

Variable	(1)	(2)	(3)	(4)
	State-Owned	Non-State-Owned	State-Owned	Non-State-Owned
	lnGreenInnovation	lnGreenInnovation	lnCite	lnCite
LnDigital	0.031 ***	0.011 ***	0.047 ***	0.005
	(0.006)	(0.004)	(0.009)	(0.005)
Control	Yes	Yes	Yes	Yes
Individual fixed effects	Yes	Yes	Yes	Yes
Industry–year fixed effects	Yes	Yes	Yes	Yes
Province–year fixed effects	Yes	Yes	Yes	Yes
Adjusted R-squared	0.718	0.623	0.803	0.745
Observations	12,556	25,889	12,559	25,907
Suset test	3.01 *		4.81 **
	(0.0827)		(0.028)

Note: Significance at the 1%, 5%, and 10% thresholds are marked ***, **, and *, respectively; the values in brackets are robust standard errors. The variables for control are the same as those in column (6) of Table 3. The coefficient difference test results were obtained based on the seemingly unrelated model test.

. They demonstrate that digital change markedly increases the quantity of green innovation in both government and private organizations, but the coefficient is lower for non-state-owned firms relative to state-owned ones. In terms of quality, digital transformation markedly enhances quality in government-owned companies, while exhibiting no substantial impact in non-state-owned enterprises. The inter-group difference test indicates significant variations in the effects of digital transformation on the quantity and quality of green innovation among different types of organizations. Compared with privately-owned enterprises, state-owned enterprises occupy a unique position in the development process with their policies, institutional advantages, capital inclination, and extensive external financing channels. In the field of industry–university–research cooperation, state-owned enterprises frequently assume a leading role (Song et al., 2022) [17]. The state-owned background has created a good green R&D innovation environment for state-owned enterprises, enabling them to attract many highly educated talents and gather rich green technology innovation resources.

4.5.2. Locational Heterogeneity

Considering the significant regional differences in China regarding economic growth, digital infrastructure, and innovation resource endowments, the influence of digital transformation on company environmental innovation may differ by location. Therefore, this study categorizes firms into two subsamples according to geographic region: the eastern region and the central and western regions.

Table 11. Results of regression of locational heterogeneity.

Variable	(1)	(2)	(3)	(4)
	East	Central and Western	East	Central and Western
	lnGreenInnovation	lnGreenInnovation	lnCite	lnCite
LnDigital	0.019 ***	0.016 **	0.017 ***	0.008
	(0.004)	(0.007)	(0.006)	(0.009)
Control	Yes	Yes	Yes	Yes
Individual fixed effects	Yes	Yes	Yes	Yes
Industry–year fixed effects	Yes	Yes	Yes	Yes
Province–year fixed effects	No	No	No	No
Adjusted R-squared	0.670	0.610	0.771	0.744
Observations	27,429	10,945	27,462	10,955
Suset test	9.7 ***		—
	0.002

Note: Significance at the 1%and 5% thresholds are marked *** and ** respectively. The values in brackets are robust standard errors. The variables for control are the same as those in column (6) of Table 3. The coefficient difference test results were obtained based on the seemingly unrelated model test. “—” indicates that the coefficients are significantly different, so no Suset test was performed

shows that the impact on enterprises in the eastern region is more significant. For firms in the central and western regions, digital transformation appears to enhance the quantity of green innovation, but the estimated coefficient is smaller than that of the eastern region, and tests for apparent irrelevance indicate a significant difference between the two. Moreover, digital transformation lacks a statistically significant influence on the quality of green innovation in the middle and western areas. This disparity may be attributed to several factors: first, the eastern region benefits from higher economic development and has accumulated diverse digital technologies and green innovation resources, creating an external environment conducive to green technological innovation; second, companies in the eastern region typically demonstrate higher degrees of digitalization and better integration of digital technologies into their operations, effectively facilitating green innovation. In contrast, the west and central areas exhibit diminished levels of economic development, which constrain improvements in the quantity and quality of green innovation. Additionally, digitalization in these regions is still at an early stage, and the integration of digital technologies with firm operations remains limited.

4.5.3. Heterogeneity of Technological Attributes

Since digital transformation requires certain technical support and resource investment, high-tech enterprises have obvious advantages over non-high-tech enterprises in digital transformation via means of their innovative resources and talent base. They can more effectively utilize innovative resources and fully leverage the benefits of digital technology, thereby improving green innovation. This paper categorizes the sample enterprises based on their classifications as companies with high technology, assigning a value of 1 to high-tech enterprises and a value of 0 to non-high-tech enterprises. Columns (1) through (4) of

Table 12. Regression results of heterogeneity of technological attributes.

Variable	(1)	(2)	(3)	(4)
	High-Tech	Non-Tech	High-Tech	Non-Tech
	lnGreenInnovation	lnGreenInnovation	lnCite	lnCite
LnDigital	0.028 ***	0.003	0.029 ***	−0.006
	(0.005)	(0.004)	(0.006)	(0.006)
Control	Yes	Yes	Yes	Yes
Individual fixed effects	Yes	Yes	Yes	Yes
Industry–year fixed effects	No	No	No	No
Province–year fixed effects	Yes	Yes	Yes	Yes
Adjusted R-squared	0.660	0.616	0.767	0.751
Observations	23,122	15,401	23,146	15,420
Suset test	367.07 ***		569.97 ***
	(0.000)		(0.000)

Note: Significance at the 1% thresholds is marked ***; the values in brackets are robust standard errors. The variables for control are the same as those in column (6) of Table 3.

exhibit the associated regression outcomes. The research results demonstrate a substantial positive correlation between the high-tech enterprises and both the quantity and quality of green innovation, with regression coefficients of 0.028 and 0.029, respectively. This signifies that for each 1% rise in digital transformation, the quantity and quality of green innovation output from high-tech firms will augment by around 0.028% and 0.029%, respectively. Conversely, the regression coefficient for non-high-tech firms failed to attain statistical significance, suggesting that the influence is not evident. The results of additional inter-group difference tests corroborated this finding, demonstrating that the influence of high-tech firms is markedly more pronounced than that of non-high-tech firms.

Liu et al. (2023) [12] used a panel threshold regression model of Chinese listed companies to investigate the consequences of digital transformation on corporate green innovation. Their findings indicate that the relationship between them is nonlinear, but rather exhibits a threshold effect. Only when digitalization reaches a certain level does a company’s innovation output significantly increase. Compared to Liu et al. (2023) [12], to analyze the effects of digital transformation on both categories of green innovation, we employed a multi-dimensional fixed-effects regression model, focusing on both the quantity and quality dimensions of innovation outcomes. We emphasized the linear connection between the two and further divided green innovation into two categories: “quantity” and “quality”, to more accurately reflect the impact between them.

Ning et al. (2023) [21] examined the roles of external environmental factors. Their empirical regression analysis of listed companies found that, under the conditions of strict environmental supervision, significant government innovation subsidies and high external media attention, digital transformation has a stronger positive impact on corporate green innovation. Shen and Tan (2022) [14] focused on internal corporate mechanisms, exploring how digital transformation can augment corporate environmental innovation by reducing information barriers, alleviating agency conflicts, and strengthening internal controls. Using questionnaire surveys and multi-level regression analysis, they demonstrated that digital technology could improve corporate internal governance structures and indirectly promote green innovation. Compared to their research, we elaborated on the information interaction effect and resource integration effect, thus enriching the research on the mechanism between them.

5. Conclusions

Drawing on the regression analysis, this study derives the findings regarding the mechanisms through which digital transformation affects corporate green innovation. (1) The findings indicate that digital transformation markedly improves both quantity and quality of corporate green innovations, with a stronger effect on the quantity. These results remain robust after conducting various robustness checks, including substituting the dependent variables, replacing the key explanatory variables, adjusting the estimation method, and employing the instrumental variable approach, indicating that Hypothesis 1, “Digital transformation can promote the incremental improvement of corporate green innovation,” is established. (2) Regarding Hypothesis 2, “Digital transformation promotes corporate green innovation by improving resource allocation efficiency,” the results of the mechanism analysis show that digital transformation increases the quantity of corporate green innovation by improving resource integration efficiency, but has no significant impact on the quality of green innovation. Hypothesis 2 holds true for the quantity of green innovation, but not for its quality. Meanwhile, regarding Hypothesis 3, “Digital transformation reduces the risk of information asymmetry in enterprises, improves the information interaction capabilities of enterprises, and motivates enterprises to carry out green innovation,” the results demonstrate that digital transformation enhances two categories of green innovation in enterprises via information interaction effects, and Hypothesis 3 is valid. (3) The heterogeneity analysis indicates that digital transformation influences both the quantity and quality of green innovation more profoundly in state-owned firms than in privately held ones, and in firms situated in the eastern region compared to those in the central and western regions. The influence of digital transformation is particularly significant in high-tech firms on both the quantity and quality of green innovation.

6. Practical Implications

This research reveals the micro-status of Chinese listed companies in promoting digitalization and innovation, offering an empirical foundation for developing policies that integrate industrial green upgrading with digital transformation at the macro level. Drawing from the study results, this paper outlines the following practical recommendations: (1) The government should encourage and support enterprises to apply technology in digital form, such as cloud computing, blockchain and big data production, management, and innovation. By formulating differentiated policies, providing special subsidies and regulatory guidance, we can promote the digitalization of enterprises to empower green innovation and promote sustainable economic development. Secondly, the government ought to augment investment in digital infrastructure, such as artificial intelligence, to narrow the regional “digital divide” and provide basic support for enterprises’ green innovations. (2) Enterprises may fully leverage digital technology to build internal and external information sharing and resource integration platforms, improve information transparency and efficiency of resource allocation, and enhance green innovation capabilities. Meanwhile, enterprises should formulate phased transformation plans based on their own characteristics, draw lessons from the practices of state-owned and high-tech enterprises, and give priority to promoting key business links of digital empowerment within the scope of their own conditions. Finally, enterprise managers can also promote the network collaboration of green innovation through digital resource integration and knowledge sharing to achieve low-carbon development of enterprises. In summary, this study offers practical guidance for governments, firms, and participants in the innovation ecosystem; highlights the main pathways and priority areas through which digital transformation fosters corporate green innovation; and provides an evidence-based foundation for promoting sustainable economic and social development.

7. Limitations of This Study and Future Research

This study still has certain limitations, and some issues could be further explored in subsequent research. First, this study primarily analyzes listed Chinese companies and does not include comparative analysis of companies in other countries. This provides insights into subsequent research. Second, owing to limitations in data availability, we concentrated on Chinese listed companies and did not include non-listed firms. Since most non-listed firms are small and micro-sized firms, which may encounter various resource limitations and external environmental pressures during digital transformation, subsequent studies may broaden the sample coverage to examine the effects of digital transformation on green innovation in small and micro enterprises. Finally, this paper fails to fully consider the strategic choices companies make in green innovation. Some companies may choose to engage in superficial or superficial green innovation activities to circumvent regulation or enhance their public image, rather than truly pursuing substantive improvements in sustainable development. Future research could focus on identifying and measuring companies’ strategic green innovation behaviors, distinguishing between substantive green innovation and strategic “greenwashing” innovation, to more accurately assess the actual contribution of digital transformation to corporate sustainable development.