The Impact of Enterprise Digital–Intelligent Transformation on Green Innovation: Empirical Evidence from Chinese Listed Companies

Abstract

As the digital economy develops and the green transition advances, digital–intelligent transformation has become a critical pathway for firms to enhance green innovation and achieve high-quality development. However, existing research lacks a systematic analysis of its role in driving green innovation. Based on panel data from Chinese A-share listed companies from 2010 to 2023, in this study, we constructed an internal capability framework from three dimensions: innovation resource allocation, human capital level, and operational management efficiency. We examined the impact of digital–intelligent transformation on green innovation and its underlying mechanisms, while incorporating the moderating role of executive green cognition. The findings indicate the following: (1) Digital–intelligent transformation significantly promotes green innovation, and this effect remains robust across instrumental variable estimation and other robustness tests. (2) The effect is more evident in state-owned enterprises, high-tech firms, and non-heavily polluting firms. (3) Mechanism analysis shows that digital–intelligent transformation promotes green innovation by optimizing resource allocation, upgrading human capital, and enhancing operational efficiency. (4) Executive green cognition strengthens the positive effect of digital–intelligent transformation on green innovation. As executive green cognition increases, the green innovation effect of digital–intelligent transformation becomes stronger. Consequently, we regard digital–intelligent transformation as an important manifestation of firms’ dynamic capability evolution in the digital context. Our study expands research on the ability of digital–intelligent transformation to affect green innovation and provides practical implications for firms promoting green and low-carbon transformation, thereby contributing to broader sustainability objectives through innovation-driven green development.

1. Introduction

Against the backdrop of intensified climate governance and growing emphasis on sustainability, green innovation is increasingly regarded as an important pathway for firms to meet environmental objectives and strengthen their long-term competitiveness. It is also critical for supporting sustainable development and facilitating the transition toward a low-carbon economy. As micro-level actors in economic activity, firms must not only reduce emissions under regulatory pressures but also move beyond passive compliance toward proactive green development through green technology R&D, process optimization, and improved resource efficiency. However, green innovation typically involves high investment, long cycles, substantial risk, and delayed returns, making it difficult for firms to sustain such activities under traditional organizational and operational models. In 2024, China issued the Opinions on Accelerating the Comprehensive Green Transformation of Economic and Social Development, which emphasizes the need to promote the deep integration of industrial digitalization, intelligent technologies, and greening. This policy provides a critical institutional backdrop for examining how digital–intelligent technologies can empower green transformation and promote sustainable corporate transformation.

Digital–intelligent technologies, represented by artificial intelligence, big data, cloud computing, the industrial Internet, and intelligent manufacturing, are being widely applied in firms’ production, management, and innovation activities. These technologies are driving firms from the stage of traditional digitalization toward a new stage of digital–intelligent transformation through the synergistic application of digital and intelligent technologies [1]. On the one hand, digital–intelligent transformation enables firms to identify green innovation opportunities more efficiently. Through intelligent analysis, algorithm optimization, and real-time coordination mechanisms, firms can break information barriers both inside and outside the organization and improve their responsiveness to market demand and external environmental changes. On the other hand, digital–intelligent transformation promotes adjustments in organizational management and innovation models, thereby strengthening organizational support for green innovation activities [2]. Compared with traditional digital transformation, digital–intelligent transformation places greater emphasis on integrating intelligent technologies into business scenarios. This helps firms optimize internal resource allocation and coordination capabilities more precisely in complex and dynamic environments, thereby maintaining long-term competitive advantages and providing a new capability foundation for the continuous promotion of green innovation [3]. Such innovation is essential for balancing environmental protection, economic performance, and social sustainability.

Building on this, we examined the impact of corporate digital–intelligent transformation on green innovation and explored its underlying mechanisms. With this study, we aim to provide practical pathways for enterprises to advance green innovation, facilitate green and low-carbon transformation, and achieve high-quality sustainable development through digital–intelligent technologies, while offering empirical evidence to inform policy formulation.

2. Literature Review

Growing attention to green development and the rapid diffusion of digital–intelligent technologies have stimulated increasing academic interest in digital–intelligent transformation and corporate green innovation. Existing research mainly examines two related issues: the drivers of corporate green innovation and the relationship between digital–intelligent transformation and green innovation.

On the one hand, research on the determinants of corporate green innovation generally considers two dimensions: external environment and internal characteristics. At the external level, factors such as government policies [4], environmental regulations [5], public oversight [6], industrial agglomeration [7], data sharing [8], and digital strategies [9] significantly influence corporate green innovation. At the internal level, studies emphasize the roles of corporate social responsibility [10], equity structure [11], executive gender [12], executive green experience [13], and digital technology innovation [14].

On the other hand, existing studies on the relationship between digital–intelligent transformation and green innovation have mostly focused on either digitalization or intelligence as a single dimension [15,16]. These studies generally suggest that digital technologies support corporate green innovation by improving information transparency [17], optimizing resource endowments [18], increasing R&D investment [19], and alleviating resource misallocation [20]. Furthermore, the embedding of intelligent technologies strengthens firms’ intelligent cognition and real-time decision-making capabilities. It also promotes the dynamic optimization of production processes and organizational operations, thereby expanding the scope for technology-enabled green development [21].

On this basis, some studies have begun to examine the comprehensive impact of digital–intelligent transformation on green innovation. From a macro perspective, city-level digital–intelligent transformation can promote corporate green collaborative innovation [22]. At the firm level, digital–intelligent transformation not only increases the number of green patents but also significantly improves green innovation efficiency through both quality improvement and scale expansion [23]. Regarding its specific mechanisms, scholars have also examined this issue from the perspectives of alleviating financing constraints [24], cultivating new quality productive forces [1], and improving executive compensation incentives [25].

Although existing studies have revealed the effect of digital–intelligent transformation on green innovation from the perspective of technology empowerment, there is still room for further research. First, most studies examine the impact of either digital transformation or intelligent application on green innovation from a single dimension. Less attention has been paid to the comprehensive effect of digital–intelligent transformation. Second, related studies mainly explain the influence paths of digital–intelligent transformation from a single aspect, such as financing constraints, external environment, or innovation investment. However, how digital–intelligent transformation affects green innovation through firms’ internal capability system remains insufficiently explained. In addition, the relationship between executives’ individual characteristics and green innovation behavior has been examined from multiple perspectives. However, whether executive green cognition influences the process through which digital–intelligent capabilities are transformed into green innovation outcomes still requires further investigation.

Given this, this study makes three contributions. First, we moved beyond the single perspective of digitalization or intelligence and examined the comprehensive impact of corporate digital–intelligent transformation on green innovation from the perspective of digital–intelligent integration. Second, we analyzed the mechanisms through which digital–intelligent transformation affects green innovation from three dimensions: resource allocation optimization, human capital upgrading, and operational efficiency improvement. We further explained how digital–intelligent transformation influences firms’ environmental perception, resource integration, and organizational coordination from the perspective of dynamic capability enhancement. Third, we incorporated executive green cognition into the analytical framework and examined how executive green cognition affects the green innovation effect of digital–intelligent transformation, thereby further explaining differences in firms’ green transformation from the perspective of managerial cognition.

3. Theoretical Analysis and Research Hypotheses

3.1. The Direct Impact of Corporate Digital–Intelligent Transformation on Green Innovation

Dynamic capability theory emphasizes that firms need to continuously sense external changes, seize market opportunities, and reconfigure internal resources in dynamic environments to maintain competitive advantages [26]. Unlike general innovation, green innovation pursues not only technological progress and economic returns but also lower resource consumption, reduced pollution emissions, and cleaner production. It therefore places higher requirements on firms’ ability to screen green technologies, process information, and identify environmental signals. As a concrete manifestation of firms’ dynamic capabilities in the digital era, digital–intelligent transformation is not merely the application of digital–intelligent technologies. It also reflects the process through which firms use data algorithms and intelligent systems to strengthen their sensing, seizing, and reconfiguring capabilities in response to external uncertainty, thereby promoting the upgrading of production and management processes [3]. From the perspective of production process optimization, digital–intelligent transformation can improve the digitalization and intelligence level of production equipment. It also promotes the intelligent and refined upgrading of production processes, thereby improving energy efficiency, reducing pollutant emissions, and creating practical conditions for the application and diffusion of green technologies [27]. From the perspective of information processing, intelligent algorithms and digital platforms can enhance firms’ efficiency in identifying external information. They help firms capture industry technology trends and changes in market demand for green products in a timely manner. This alleviates information asymmetry in green technological innovation [28] and improves the economic feasibility of green innovation. At the same time, based on big data analysis, firms can more accurately guide innovation resources toward green fields and accelerate green innovation activities. Based on the above analysis, the following hypothesis is proposed:

H1: 

Corporate digital–intelligent transformation significantly promotes green innovation.

3.2. Mechanisms Through Which Corporate Digital–Intelligent Transformation Drives Green Innovation

Green innovation activities depend not only on technological investment, but also on firms’ continuous support in resource allocation, knowledge absorption, and organizational operations. In this subsection, we analyze the mechanisms through which digital–intelligent transformation affects green innovation from three dimensions: innovation resource allocation, human capital upgrading, and operational efficiency improvement. Among them, innovation resource allocation reflects firms’ ability to support green innovation activities with resources. Human capital upgrading captures firms’ ability to learn and absorb green technologies and digital–intelligent technologies. Operational efficiency improvement reflects firms’ organizational implementation capability in embedding green technologies into production and operational processes.

3.2.1. Resource Allocation Optimization Mechanism

Resource-based theory posits that sustained and effective resource investment is a critical prerequisite for firms to engage in innovation activities. Compared with general innovation, green innovation involves greater uncertainty and a longer R&D cycle. It therefore places higher requirements on the continuity and efficiency of firms’ resource input and allocation. Innovation resource allocation capability affects not only the resource support for green technology R&D, but also whether firms can effectively allocate limited resources to key areas such as emission reduction, green processes, and environmental technology development [23]. Financial resources, as core innovation resources, determine the quality and direction of green output. They also affect the continuity of green innovation strategies [29]. Therefore, resource allocation optimization is an important supporting condition for firms to carry out green innovation.

Through digital–intelligent transformation, firms can improve information processing efficiency and data analysis capabilities. This helps optimize resource allocation decisions and accelerate the integration of innovation factors. On the one hand, digital technologies improve the transparency of internal R&D, production, financial, and market information. They can also reduce information dispersion and resource misallocation under traditional management models, enabling firms to identify key links and resource needs in green R&D more accurately. On the other hand, intelligent algorithms and data analytics tools help firms assess the technical feasibility, market prospects, and potential risks of green innovation projects, thereby enhancing the scientific rigor of R&D investment decisions [30]. Therefore, H2 is proposed:

H2: 

Corporate digital–intelligent transformation promotes green innovation by optimizing resource allocation.

3.2.2. Human Capital Upgrading Mechanism

According to endogenous growth theory, human capital serves as a critical source of firms’ technological progress and innovation capability. Green innovation usually involves the interdisciplinary integration of environmental technologies, digital technologies, and production processes. It therefore places higher requirements on employees’ professional skills and knowledge absorption capacity. Whether firms possess a high level of human capital may directly affect their ability to identify, absorb, and transform green technological knowledge. A higher level of human capital helps firms accelerate the learning and diffusion of green technological knowledge. It also determines whether firms can effectively identify, absorb, and transform green technological achievements [31]. Through substitution, complementarity, and accumulation effects, digital–intelligent transformation propels a systematic evolution in corporate talent structures. First, the substitution effect arises from the adoption of intelligent equipment, such as AI and automated machinery, which replaces routine and standardized tasks, reducing firms’ reliance on traditional low-skilled labor [32]. Second, the complementarity effect is reflected in the increased demand for highly skilled and multidisciplinary talent, prompting firms to adjust their human capital structure toward greater specialization and professionalization [33]. Finally, considering cumulative effects, the application of digital–intelligent systems accelerates internal knowledge accumulation within firms. It helps employees master digital–intelligent tools and practical skills more quickly, thereby continuously improving the overall level of firms’ human capital. Accordingly, H3 is proposed:

H3: 

Corporate digital–intelligent transformation enhances green innovation by upgrading human capital.

3.2.3. Operational Efficiency Enhancement Mechanism

Unlike resource allocation optimization, which emphasizes resource input, and human capital upgrading, which focuses on knowledge absorption and transformation, operational efficiency improvement focuses more on improvements in firms’ internal organizational operations, management coordination, and execution efficiency. Digital–intelligent transformation can promote systematic changes in business processes, organizational structures, and management methods. It makes organizational structures flatter, reduces information transmission layers [34], and improves management’s ability to perceive production operations and environmental risks. At the same time, digital–intelligent systems can strengthen internal supervision and process control. They enable shareholders and managers to obtain key operating indicators and resource flow information in a timelier manner. This reduces agency costs caused by information asymmetry, limits managerial opportunistic behavior, and further improves overall operational management efficiency [35]. Improved operational efficiency helps reduce organizational friction costs in the process from R&D design to production application [36]. It also enables green technological solutions to be more smoothly embedded in energy-saving management and low-carbon processes, thereby promoting firms’ green innovation output. Accordingly, H4 is proposed:

H4: 

Corporate digital–intelligent transformation enhances green innovation by improving operational efficiency.

3.3. The Moderating Role of Executive Green Cognition

Upper echelons theory posits that senior executives’ perceptions influence corporate strategic decisions and resource allocation. Executive green cognition refers to the environmental responsibility awareness and green development orientation that executives formulate at the strategic level, which are reflected in corporate green innovation practices through formal disclosures or strategic decisions [37]. As a critical determinant of corporate strategic decision-making, executive cognition plays a key role in firms’ green transformation. On the one hand, executives’ green cognition facilitates the integration of digital–intelligent transformation into the firm’s green development strategy, thereby strengthening strategic guidance and resource allocation throughout the green innovation process. On the other hand, executives with strong green cognition are better able to perceive market green demand and environmental regulatory signals, optimize investment decisions in green innovation, and enhance the coordination efficiency between digital–intelligent technologies and green resources [38]. Furthermore, executives can turn environmental limitations and competitive threats into motivators for green innovation. This conversion, in turn, facilitates the adoption of digital–intelligent technologies in sustainable manufacturing and management practices [39]. Accordingly, H5 is proposed:

H5: 

Executive green cognition positively moderates the effect of corporate digital–intelligent transformation on green innovation. Specifically, greater executive green cognition amplifies the positive impact of digital–intelligent transformation on green innovation.

4. Research Design

4.1. Model Specification

4.1.1. Baseline Regression Model

To examine the effect of corporate digital–intelligent transformation on green innovation, in this study we specified the baseline regression model:

$${GTI}_{it}={\alpha }_0+{\alpha }_1{DI}_{it}+{\alpha }_2{Controls}_{it}+Firm+Year+{\epsilon }_{it}$$

(1)

where i and t denote firm and year, respectively; GTI represents the level of green innovation; DI denotes digital–intelligent transformation; Controls represents a set of control variables; and ε_it is the idiosyncratic error term. We incorporated both firm-level and year-level fixed effects, denoted as Firm and Year.

4.1.2. Mechanism Testing Model

In this study, we investigated the specific pathways through which corporate digital–intelligent transformation affects green innovation, focusing on three mechanisms: resource allocation optimization, human capital upgrading, and operational efficiency enhancement. Following Yang Haochang et al. [23], we employed stepwise regression for empirical analysis, supplemented by Bootstrap methods for statistical inference. The mechanism testing model is specified as follows:

$$M_{it}={\beta }_0+{\beta }_1{DI}_{it}+\beta {Controls}_{it}+Firm+Year+{\epsilon }_{it}$$

(2)

$${GTI}_{it}={\gamma }_0+{\gamma }_1{DI}_{it}+{\gamma }_2{M}_{it}+{\gamma }_3{Controls}_{it}+Firm+Year+{\epsilon }_{it}$$

(3)

Here, M represents the mechanism variables, including innovation resource allocation, human capital, and operational management efficiency; all other variables are consistent with those in Model (1).

4.1.3. Model for Testing the Moderating Effect

In this study, we incorporated an interaction term between corporate digital–intelligent transformation and executive green cognition into the baseline regression model to examine the moderating role of EGC in the relationship between digital–intelligent transformation and green innovation. The model is specified as follows:

$${GTI}_{it}={\delta }_0+{\delta }_1{DI}_{it}+{\delta }_2{EGC}_{it}+{\delta }_3{{DI}_{it}\times EGC}_{it}+{\delta }_4{Controls}_{it}+Firm+Year+{\epsilon }_{it}$$

(4)

Here, EGC represents the moderator variable, reflecting executives’ green cognition, and DI × EGC denotes the interaction between digital–intelligent transformation and EGC; all other variables are consistent with those in Model (1).

4.2. Selection of Variables

4.2.1. Dependent Variable

Green Innovation. Following the methodology of Li Qingyuan and Xiao Zhehua [40], corporate green innovation (GTI) was measured as the natural logarithm of the total number of green patent applications filed by listed companies each year, plus one. Such a measure encompasses two categories: green invention patent applications and green utility model patent applications.

4.2.2. Key Explanatory Variable

Digital–Intelligent Transformation. We measured the level of corporate digital–intelligent transformation using annual report text analysis. Following the methods of Wu Fei et al. [41] and Liu Lingbing et al. [42], 68 keywords related to digital–intelligent transformation were categorized into three dimensions: digital technology infrastructure, intelligent technology applications, and digital–intelligent integration and empowerment. During text processing, Python (version 3.13.5) with the Jieba tokenization tool was used for text recognition and segmentation of annual reports, and synonyms and alternative forms of keywords were standardized and merged. To reduce the influence of differences in report length and extreme keyword frequencies across firms, the digital–intelligent transformation index was constructed as ln(1 + frequency) of relevant keywords. Detailed text processing procedures are provided in Appendix A.

4.2.3. Mechanism Variables

Resource Allocation Optimization Mechanism: Innovation Resource Allocation. Following Fang Wenlong et al. [43] and CANACE et al. [44], innovation resource allocation reflects a firm’s tendency to allocate and invest in innovation resources, measured by R&D intensity (RD). R&D intensity captures a company’s commitment to innovation activities and its allocation preferences, serving as an important prerequisite for green technology R&D. Specifically, it is calculated as the ratio of R&D expenditure to total assets.

Human Capital Upgrading Mechanism: Human Capital Level. Following Liu Qiren and Zhao Can [45] and Niu Zhiwei et al. [46], we measured human capital across two dimensions, skill structure and educational structure, reflecting the firm’s human capital base and technological absorption capacity. Skill level (Labor_Skill) is defined as the proportion of technical staff in the total workforce. Educational level (Labor_Edu) is calculated as the percentage of the workforce holding a Bachelor’s degree or higher. Together, these indicators reflect a firm’s knowledge base and its capacity to learn innovative technologies, providing essential human capital for absorbing advanced digital–intelligent technologies and applying them to green innovation.

Operational Efficiency Enhancement Mechanism: Operational Management Efficiency. Green innovation occurs not only in R&D but also within operational processes, such as production workflows, energy-saving management, and pollution control. Asset turnover ratio (ATO) reflects a firm’s operational efficiency and managerial coordination capabilities, illustrating the organizational foundation for translating green technology solutions into practice. Accordingly, following Li Wengui et al. [47], we utilized the asset turnover ratio as an indicator of management efficiency.

4.2.4. Moderating Variable

Executive Green Cognition. Text analysis has been widely used in organizational and strategic management research, particularly in studies of managerial perceptions and strategic attention, and is well suited for longitudinal data [48]. Corporate annual reports provide an observable basis for capturing managerial perceptions [49]. Following Song Jing et al. [39] and Li Yabing et al. [50], we selected 19 keywords across three dimensions of executive green cognition: cognition of green competitive advantage, cognition of corporate social responsibility, and cognition of external environmental pressure. By counting the frequency of these keywords in the annual reports of listed companies, we measured the level of executive green cognition (EGC), which reflects executives’ green strategic orientation and managerial attention to environmental issues as disclosed in formal corporate reports. The detailed keyword dictionary and text processing procedures are provided in Appendix A.

4.2.5. Control Variables

Drawing on prior studies [51,52], we included a set of control variables: firm size (Size), proxied by the natural logarithm of total assets; firm value (TobinQ), proxied by Tobin’s Q; return on equity (ROE), defined as the ratio of net profit to shareholders’ equity; dual role (Dual), coded as 1 if the chairman and general manager are the same person and 0 otherwise; and firm age (FirmAge), calculated as ln(1 + establishment duration in years). In addition, the regression model incorporates firm and year fixed effects. Firm fixed effects capture time-invariant firm characteristics, such as governance structure, green investment preferences, and regional attributes. Year fixed effects account for macroeconomic policies and fluctuations in the business cycle.

4.3. Data Sources and Descriptive Statistics

In this study, we used Chinese A-share listed companies on the Shanghai and Shenzhen Stock Exchanges from 2010 to 2023 as the research sample. Corporate green patent data were obtained from the China Research Data Service Platform (CNRDS); annual reports were collected from the CNINFO website. Furthermore, the CSMAR and Wind databases provided essential firm-level characteristics, financial figures, and industry categorization.

To ensure data validity, the following steps were undertaken: (1) exclusion of listed companies in the finance and insurance sectors; (2) exclusion of firms designated as ST, *ST, or PT during the study period; (3) exclusion of observations with missing values for key variables; and (4) winsorization of all continuous variables at the 1st and 99th percentiles. The resulting unbalanced panel dataset consists of 24,189 observations covering 2885 listed companies. Descriptive statistics for the main variables are presented in

Table 1. Descriptive statistics of main variables.

Variables	N	Mean	SD	Min	Median	Max
GTI	24,189	0.9528	1.2337	0.0000	0.6931	5.3181
DI	24,189	1.4576	1.4092	0.0000	1.0986	6.2246
Size	24,189	22.2158	1.3564	19.5628	21.9610	26.4523
TobinQ	24,189	2.0305	1.3166	0.7888	1.6229	16.6472
ROE	24,189	0.0689	0.1246	−0.9616	0.0775	0.4140
Dual	24,189	0.3070	0.4612	0.0000	0.0000	1.0000
FirmAge	24,189	2.9243	0.3326	1.0986	2.9444	3.6376
RD	20,890	2.6017	2.6248	0.0000	2.0777	53.8365
Labor_Skill	23,636	0.2404	0.2004	0.0002	0.1719	1.0000
Labor_Edu	24,189	0.3161	0.2378	0.0000	0.2542	1.0000
ATO	24,188	0.5855	0.3932	0.0528	0.4950	2.6456
EGC	24,189	3.3064	4.4180	0.0000	2.0000	22.0000

.

5. Empirical Results and Analysis

5.1. Baseline Regression Results

Table 2. Results of the baseline regression.

Variables	(1) GTI	(2) GTI
DI	0.0726 *** (0.0105)	0.0386 *** (0.0100)
Size	-	0.3744 *** (0.0258)
TobinQ	-	0.0042 (0.0062)
ROE	-	−0.0005 (0.0528)
Dual	-	0.0069 (0.0204)
FirmAge	-	0.1228 (0.1378)
_cons	0.8469 *** (0.0153)	−7.7882 *** (0.6226)
Firm	YES	YES
Year	YES	YES
N	24,189	24,189
R2	0.7668	0.7783

Note: Robust standard errors clustered at the firm level are reported in parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

summarizes the baseline findings regarding the impact of corporate digital–intelligent transformation on green innovation. As shown in Table 2, we report the results both without (Column 1) and with (Column 2) a vector of control variables. The results show that the coefficient of DI is positive and statistically significant at the 1% level in both specifications, suggesting that DI transformation significantly promotes firms’ green innovation (GTI).

In terms of economic magnitude, the estimated coefficient of DI in Column (2) is 0.0386. Given the standard deviation of DI of 1.4092, a one-standard-deviation increase in digital–intelligent transformation is associated with an increase of approximately 5.44% in green innovation (0.0386 × 1.4092 × 100%). This indicates that DI can improve the efficiency of green innovation investment and technology application. It also provides more favorable internal support for green innovation activities. These findings have practical implications for firms seeking to improve green innovation and achieve high-quality economic development. Accordingly, H1 is supported.

5.2. Endogeneity Test

In this study, we employed an instrumental variable approach to address potential endogeneity in the baseline regression results. Following Lewbel [53] and Meng Hao et al. [54], we adopted an IV calculated as the cubed deviation of a company’s digital–intelligent transformation from its industry-year mean. Technological diffusion and demonstration effects within industries imply that a firm’s deviation from the industry-year average level of digital–intelligent transformation is closely associated with its own transformation level. Therefore, the instrument satisfies the relevance condition. Regarding the exclusion restriction, this deviation mainly captures the extent to which a firm differs from its industry peers in digital–intelligent transformation, rather than its green R&D investment, environmental governance expenditure, or green patent output. Thus, the instrument is expected to affect green innovation primarily through firms’ own digital–intelligent transformation, rather than directly through green innovation activities or environmental governance strategies. In addition, the cubic transformation was introduced to strengthen the non-linear variation in the instrument while preserving its exogenous component relative to firm-level green innovation outcomes, which further supports the validity of the instrumental variable.

Table 3. Instrumental variable estimation results.

Variables	(1)	(2)
	First Stage: DI	Second Stage: GTI
IV	0.1343 *** (0.0063)	-
DI	-	0.0469 *** (0.0166)
Controls	YES	YES
Firm	YES	YES
Year	YES	YES
Kleibergen–Paap rk LM statistic	437.294 ***
Kleibergen–Paap Wald rk F statistic	453.479 [16.38]
Cragg–Donald Wald F statistic	1.3 × 104
N	24,189	24,189
R2	-	0.0541

Note: Robust standard errors clustered at the firm level are reported in parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

reports the instrumental variable regression results. The first-stage regression shows that the coefficient of IV is positive and statistically significant at the 1% level, indicating that the instrument is strongly correlated with firms’ digital–intelligent transformation. Moreover, the Kleibergen–Paap rk LM statistic is significant at the 1% level, rejecting the null hypothesis of underidentification. The Kleibergen–Paap Wald rk F statistic is 453.479, which is substantially higher than the Stock–Yogo critical value of 16.38 at the 10% maximal IV size level, suggesting that weak-instrument concerns are unlikely to arise. These diagnostic tests support the validity of the selected IV. The second-stage regression results further reflect that, after addressing potential endogeneity using the instrumental variable approach, the coefficient of DI remains positive and statistically significant. This confirms the robustness of the core conclusion that digital–intelligent transformation promotes corporate green innovation.

5.3. Robustness Tests

5.3.1. Alternative Dependent Variable

Following Xu Jia et al. [55], we replaced the total number of green patents with the number of green invention patent applications and the number of green utility model patent applications, respectively, to test the robustness of the baseline results. Specifically, two alternative dependent variables were constructed: Green_inv, measured as ln(green invention patent applications + 1), and Green_uma, measured as ln(green utility model patent applications + 1).

Table 4. Results of robustness tests.

Variables	Alternative Dependent Variable		Lagged Explanatory Variable	Winsorization	PSM Matching	Excluding Policy Shocks	High-Dimensional FE
	(1) Green_inv	(2) Green_uma	(3) GTI	(4) GTI	(5) GTI	(6) GTI	(7) GTI
DI	0.0452 *** (0.0092)	0.0150 * (0.0085)	-	0.0339 *** (0.0093)	0.0445 *** (0.0110)	0.0355 *** (0.0107)	0.0412 *** (0.0097)
L.DI	-	-	0.0417 *** (0.0102)	-	-	-	-
Controls	YES	YES	YES	YES	YES	YES	YES
Firm	YES	YES	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES	YES
N	24,189	24,189	21,296	24,189	19,296	21,564	24,189
R2	0.7528	0.7376	0.7912	0.7576	0.7925	0.7854	0.7814

Note: Robust standard errors clustered at the firm level are reported in parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

presents the results in Columns (1) and (2). Regardless of which alternative measure of green innovation is used, the coefficient of DI remains positive and statistically significant. Moreover, the coefficient for green invention patents is larger and more statistically significant, suggesting that digital–intelligent transformation plays a more pronounced role in promoting high-quality green innovation.

5.3.2. Lagged Core Explanatory Variable

To mitigate potential reverse causality, we further re-estimated the baseline model using the one-period lagged value of digital–intelligent transformation. As reported in Column (3) of Table 4, the coefficient of L.DI is 0.0417 and statistically significant at the 1% level. This result indicates that digital–intelligent transformation continues to significantly promote corporate green innovation after reverse causality concerns are partially addressed, thereby further confirming the robustness of the baseline conclusion.

5.3.3. Winsorization

To further examine whether the baseline results are sensitive to the treatment of extreme values, we re-estimated the baseline model after winsorizing all continuous variables at the 5th and 95th percentiles. As reported in Column (4) of Table 4, the coefficient of DI remains positive and statistically significant at the 1% level. This finding suggests that the baseline results are not driven by extreme observations.

5.3.4. Propensity Score Matching

To further address potential selection concerns, we performed an additional robustness check using propensity score matching (PSM), following Li Jinchang et al. [56]. Firms were divided into high and low digital–intelligent transformation groups based on the annual industry median of digital–intelligent transformation. Subsequently, we performed 1:1 nearest-neighbor matching with a caliper width set at 0.01. The balance tests show that the mean differences in covariates decrease substantially after matching, indicating satisfactory matching quality. Detailed balance test results are reported in Appendix B. Using the matched sample, we re-estimated the baseline regression. As reported in Column (5) of Table 4, the coefficient of DI remains positive and statistically significant at the 1% level. This result implies that, after mitigating sample selection bias, the positive effect of digital–intelligent transformation on green innovation remains robust.

5.3.5. Excluding Contemporaneous Policy Shocks

To rule out the potential influence of major contemporaneous policy changes on the baseline results, we further excluded observations from 2015 and 2016 and re-estimated the baseline model, following Wu Fei et al. [41]. This time window was selected because the newly revised Environmental Protection Law of the People’s Republic of China came into effect in 2015. At the early stage of policy implementation, the law may have exerted a strong influence on firms’ green innovation behavior, and firms’ responses to changes in environmental regulation may also involve a short-term lag. Therefore, excluding observations from 2015 and 2016 helps us to examine whether the main findings are affected by major environmental regulatory changes and their short-term lagged effects. As reported in Column (6) of Table 4, the coefficient of DI remains positive and statistically significant, indicating that the core conclusion of this study remains robust.

5.3.6. Controlling for High-Dimensional Fixed Effects

To further alleviate omitted-variable bias and account for unobservable industry-level factors, we augmented the baseline regression model by incorporating industry fixed effects. This specification captures time-invariant unobservable characteristics at the industry level, such as technological trajectories, regulatory environments, and resource dependence. As reported in Column (7) of Table 4, the coefficient of DI remains positive and statistically significant at the 1% level, further reinforcing the reliability of the empirical findings.

6. Further Analysis

6.1. Mechanism Tests

6.1.1. Resource Allocation Optimization Mechanism

Innovation resource allocation was used as the mechanism variable for resource allocation optimization. The results are reported in Columns (1) and (2) of

Table 5. Results of mechanism tests.

Variables	Resource Allocation Optimization		Human Capital Upgrading				Operational Efficiency Enhancement
	(1) RD	(2) GTI	(3) Labor_Edu	(4) GTI	(5) Labor_Skill	(6) GTI	(7) ATO	(8) GTI
DI	0.0777 *** (0.0188)	0.0386 *** (0.0102)	0.0041 * (0.0019)	0.0371 *** (0.0099)	0.0043 *** (0.0016)	0.0389 *** (0.0099)	0.0096 ** (0.0040)	0.0378 *** (0.0100)
RD	-	0.0341 *** (0.0057)	-	-	-	-	-	-
Labor_Edu	-	-	-	0.3569 *** (0.0916)	-	-	-	-
Labor_Skill	-	-	-	-	-	0.2859 *** (0.0918)	-	-
ATO	-	-	-	-	-	-	-	0.0779 * (0.0453)
Controls	YES	YES	YES	YES	YES	YES	YES	YES
Firm	YES	YES	YES	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES	YES	YES
N	20,870	20,870	24,189	24,189	23,630	23,630	24,188	24,188
R2	0.8439	0.7950	0.8788	0.7789	0.8676	0.7808	0.8107	0.7784

Note: Robust standard errors clustered at the firm level are reported in parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

. Column (1) shows that the coefficient of DI is positive and statistically significant at the 1% level, suggesting that digital–intelligent transformation significantly improves firms’ innovation resource allocation (RD). Column (2) further shows that the coefficients of both DI and innovation resource allocation (RD) are positive and statistically significant at the 1% level. This suggests that DI indirectly promotes green innovation by optimizing innovation resource allocation. These findings indicate that digital–intelligent transformation helps improve firms’ resource coordination capabilities in the green innovation process. It therefore strengthens the continuous support capacity for green innovation activities. The mechanism of resource allocation optimization is verified, and H2 is supported.

6.1.2. Human Capital Upgrading Mechanism

Human capital level was used as the mechanism variable for human capital upgrading. It is measured from two dimensions: skill structure and educational structure. Specifically, Labor_Skill reflects the skill level of human capital, while Labor_Edu represents educational attainment. The results are reported in Columns (3)–(6) of Table 5. The results show that the coefficients of DI are positive and statistically significant, which indicates that digital–intelligent transformation significantly improves both the skill level and educational attainment of human capital. Columns (4) and (6) further show that the coefficients of DI, employee skill level (Labor_Skill), and educational attainment (Labor_Edu) are all positive and statistically significant at the 1% level. This suggests that DI indirectly promotes green innovation by upgrading human capital. These findings indicate that digital–intelligent transformation enhances firms’ ability to learn and absorb green technologies and digital–intelligent technologies. It therefore provides knowledge and capability support for green innovation activities. The mechanism of human capital upgrading is verified, and H3 is supported.

6.1.3. Operational Efficiency Enhancement Mechanism

Operational management efficiency was used as the mechanism variable for operational efficiency improvement. The results are reported in Columns (7) and (8) of Table 5. Column (7) shows that the coefficient of digital–intelligent transformation (DI) is significantly positive at the 5% level. This indicates that DI significantly improves operational management efficiency (ATO). Column (8) further shows that the coefficients of DI and ATO are significant at least at the 10% level. This indicates that DI indirectly promotes green innovation by improving operational efficiency. These results indicate that digital–intelligent transformation helps enhance organizational coordination and process operation capabilities in green innovation activities. It therefore improves the implementation efficiency of green innovation. The mechanism of operational efficiency improvement is verified, and H4 is supported.

6.1.4. Bootstrap Mediation Tests

To further address the limitations of the traditional stepwise regression approach in identifying indirect effects, we followed Guo Jinying et al. [1] and conducted supplementary Bootstrap mediation tests. At the 95% confidence level, the empirical distributions of the indirect effects were constructed through 1000 repeated resamplings. As reported in

Table 6. Results of bootstrap mediation tests.

Variables	Effect Type	Coefficient	Standard Error	Z-Value	95% CI
RD	Indirect Effect	0.00265 ***	0.00063	4.19	[0.00141, 0.00389]
	Direct Effect	0.03859 ***	0.00833	4.64	[0.02228, 0.05491]
Labor_Edu	Indirect Effect	0.00146 ***	0.00048	3.06	[0.00053, 0.00240]
	Direct Effect	0.03711 ***	0.00704	5.27	[0.02332, 0.05090]
Labor_Skill	Indirect Effect	0.00123 ***	0.00040	3.08	[0.00045, 0.00202]
	Direct Effect	0.03893 ***	0.00731	5.32	[0.02460, 0.05325]
ATO	Indirect Effect	0.00075 **	0.00032	2.30	[0.00011, 0.00138]
	Direct Effect	0.03782 ***	0.00715	5.29	[0.02381, 0.05184]

Note: ***, **, and * denote significance levels of 1%, 5%, and 10%, respectively.

, the confidence intervals for the indirect effects by innovation resource allocation (RD), human capital (Labor_Edu and Labor_Skill), and operational management efficiency (ATO) do not include zero. The results indicate that these indirect effects are statistically significant and further support the proposed internal capability transmission mechanisms. In summary, digital–intelligent transformation promotes green innovation via three pathways: resource allocation optimization, human capital upgrading, and operational efficiency enhancement.

6.2. Moderating Effect Analysis

Based on the preceding analysis, executive green cognition may moderate the relationship between digital–intelligent transformation and green innovation. In this subsection, we empirically tested this moderating role. As reported in

Table 7. Moderating effect results.

Variables	(1) GTI
DI	0.0197 * (0.0113)
EGC	0.0010 (0.0032)
DI × EGC	0.0057 *** (0.0017)
Controls	YES
Firm	YES
Year	YES
N	24,189
R2	0.7788

Note: Robust standard errors clustered at the firm level are reported in parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

, the coefficient of the interaction term between digital–intelligent transformation and executive green cognition (DI × EGC) is significantly positive at the 1% level. This suggests that EGC positively moderates the relationship between digital–intelligent transformation and green innovation. Therefore, H5 is supported.

To further understand the economic significance of executive green cognition, we followed the approach of Busenbark et al. [57]. Based on the interaction term model, we calculated the conditional marginal effect of digital–intelligent transformation on green innovation at different levels of executive green cognition, as follows:

$${\displaystyle \frac{\partial GTI}{\partial DI}}={\delta }_1+{\delta }_{3}EGC$$

Considering the evident skewness of the executive green cognition variable, we used its 25th percentile (EGC = 0), median (EGC = 2), and 75th percentile (EGC = 4) to represent low, medium, and high levels, respectively. The results show that when EGC is at a low level, the marginal effect of digital–intelligent transformation on green innovation is 0.0197. When EGC is at a medium level, the marginal effect increases to 0.0311. When EGC is at a high level, it further rises to 0.0425. These results indicate that the positive effect of digital–intelligent transformation on green innovation becomes stronger as executive green cognition increases. This means that the green innovation effect of digital–intelligent transformation depends not only on firms’ investment in digital–intelligent technologies but also on management’s cognition of green development. Management with stronger green cognition is more likely to strengthen the green orientation in corporate governance, thereby more effectively translating the advantages of digital–intelligent transformation into green innovation outcomes.

6.3. Heterogeneity Analysis

6.3.1. Ownership Heterogeneity

Because firms differ in resource endowments and policy mandates, the strategies and outcomes of digital–intelligent transformation may vary across ownership structures. As important participants in the implementation of national green development strategies, state-owned enterprises (SOEs) may have advantages in green innovation resource inputs and policy responsiveness. Therefore, the positive effect of DI transformation on green innovation may be more evident among SOEs. In this research, we grouped the firms into SOEs and non-SOEs based on their ownership nature, and the results are reported in Columns (1) and (2) of

Table 8. Heterogeneity analysis results.

Variables	Ownership		Technological Level		Industry Pollution Level
	(1) State-Owned	(2) Non-State-Owned	(3) High-Tech	(4) Low-Tech	(5) Heavily Polluting	(6) Non-Heavily Polluting
DI	0.0646 *** (0.0184)	0.0298 *** (0.0113)	0.0371 *** (0.0123)	0.0376 ** (0.0151)	0.0338 (0.0265)	0.0476 *** (0.0104)
Controls	YES	YES	YES	YES	YES	YES
Firm	YES	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES
N	8886	15,303	13,443	10,727	5143	19,032
R2	0.8127	0.7440	0.7908	0.7727	0.7554	0.7896

Note: Robust standard errors clustered at the firm level are reported in parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

.

The coefficient of digital–intelligent transformation is positive and statistically significant for both SOEs and non-SOEs. Moreover, the estimated coefficient is relatively larger in the SOE subsample, suggesting that the green innovation effect of DI is more evident among SOEs. This pattern is consistent with the institutional and resource advantages of SOEs. On the one hand, SOEs are more likely to obtain policy support, such as green subsidies and pilot demonstration projects. As ESG-related assessments are increasingly incorporated into the supervision system for state-owned assets, SOEs’ internal motivation to engage in green innovation may be further strengthened. On the other hand, SOEs generally have relatively stable financial, human, and technological resources, enabling them to better cope with the risks associated with digital–intelligent transformation and to leverage digital–intelligent technologies to promote green innovation.

6.3.2. Technological Level Heterogeneity

Firms with different technological levels may differ in their technology absorption capacity, R&D intensity, and efficiency in converting innovation inputs into outputs. Consequently, the effect of digital–intelligent transformation on green innovation may also vary across firms with different technological levels. In this study, we divided the sample into high-tech and low-tech firms for subgroup regressions, and the results are reported in Columns (3) and (4) of Table 8. The regression results show that the coefficients of digital–intelligent transformation are positive and statistically significant in both the high-tech and low-tech firm subsamples. Notably, the coefficient is statistically significant at the 1% level in the high-tech firm subsample and at the 5% level in the low-tech firm subsample. These results suggest that digital–intelligent transformation is positively associated with green innovation among firms with different technological levels, with the effect appearing more evident among high-tech firms. This pattern may be related to high-tech firms’ technology absorption capacity and established R&D systems. High-tech firms generally have a stronger foundation for applying digital–intelligent technologies and more accumulated innovation resources, which enables them to use data-driven resources more efficiently in green R&D. In addition, their relatively mature R&D systems may help them translate the benefits of digital–intelligent transformation into green innovation outcomes more effectively, enabling digital–intelligent technologies to support green innovation.

6.3.3. Industry Pollution Level Heterogeneity

Firms in industries with different pollution levels may face different environmental constraints and resource endowments, which may lead to variation in the green innovation effect of digital–intelligent transformation. Based on the secondary industry classification standard in the Guidelines for Industry Classification of Listed Companies, we divided the sample into heavily polluting and non-heavily polluting firms for subgroup regressions. As shown in Columns (5) and (6) of Table 8, the coefficient of digital–intelligent transformation is positive and statistically significant in the subsample of non-heavily polluting firms, whereas it is not statistically significant in the subsample of heavily polluting firms. This suggests that the green innovation effect of digital–intelligent transformation is more evident among non-heavily polluting firms. One possible explanation is that heavily polluting firms tend to rely more on energy-intensive production models and face relatively high transformation costs, which may constrain the green innovation effect of digital–intelligent transformation. In addition, many heavily polluting firms in China are concentrated in traditional industrial sectors and are often characterized by weak technological accumulation and limited financing channels [58]. Under more urgent environmental compliance pressures, the role of DI transformation in promoting GTI among these firms may be relatively limited.

7. Conclusions and Implications

7.1. Conclusions

Based on data from Chinese A-share listed companies on the Shanghai and Shenzhen Stock Exchanges from 2010 to 2023, in this study, we systematically examined the effect of corporate digital–intelligent transformation on green innovation and its underlying mechanisms. The results show that digital–intelligent transformation significantly promotes green innovation. It does so mainly through three paths: resource allocation optimization, human capital upgrading, and operational efficiency improvement. Executive green cognition strengthens the positive effect of digital–intelligent transformation on green innovation. As executive green cognition increases, the green innovation effect of digital–intelligent transformation becomes stronger. In addition, heterogeneity analysis shows that this promoting effect differs across enterprises, being more pronounced in state-owned, high-tech, and non-heavily polluting firms. The findings suggest that the role of digital–intelligent transformation is not limited to the application of digital technologies. More importantly, it promotes the shift in firms’ green innovation models from factor- to capability-driven development. Green innovation depends not only on R&D investment but also on firms’ abilities to coordinate resources, knowledge, and organizational operations. At the same time, corporate governance orientation and executive green cognition also play important roles. These findings also suggest that digital–intelligent transformation can contribute to sustainability by promoting green innovation, supporting low-carbon development, and improving firms’ capacity to balance environmental and economic objectives.

Based on the above findings, we propose the following policy recommendations from the perspectives of both the government and enterprises.

At the government level, the institutional environment for the coordinated development of digital–intelligent transformation and green development should be further improved to provide stable and long-term external support for corporate green innovation. First, the government should strengthen the construction of green-oriented digital–intelligent infrastructure and public data platforms. This would reduce firms’ costs of accessing data, technologies, and green innovation resources, while encouraging the wider application of digital–intelligent technologies in green R&D, energy-saving transformation, and green process optimization. Moreover, supporting policies related to green finance, R&D subsidies, and talent support should be further improved to alleviate long-term resource constraints in corporate green innovation. In addition, the government should implement differentiated policies based on firms’ resource endowments and industry characteristics. Greater support should be provided for low-tech and heavily polluting firms to promote digital–intelligent upgrading and green process transformation, thereby gradually enhancing their green innovation capabilities.

At the enterprise level, firms should not regard digital–intelligent transformation merely as a technological tool for improving operational efficiency. More attention should also be paid to the development of green innovation capabilities. Firms should continuously strengthen their support capacity for green innovation by focusing on key areas such as resource allocation, talent cultivation, and organizational operations. This includes optimizing the allocation of green R&D resources, strengthening the cultivation of interdisciplinary talent, and promoting the optimization of organizational processes and management models. Meanwhile, corporate management should strengthen its cognition of green development and attach greater importance to green innovation goals in corporate governance. This would improve the efficiency with which the advantages of digital–intelligent transformation are translated into green innovation outcomes. Different types of firms should also formulate differentiated transformation paths based on their own resource endowments. Firms with stronger resource and technological foundations should further deepen the integration of digital–intelligent technologies with green R&D. In contrast, firms with relatively weak capability foundations should prioritize digital–intelligent upgrading and green process transformation in key production links.

7.2. Limitations and Future Research

Although this study has elucidated the relationship between enterprise digital–intelligent transformation and green innovation, several limitations remain and provide directions for future research. First, the sample of this study is limited to Chinese A-share listed companies. Therefore, the applicability of the findings to other institutional and regulatory contexts requires further examination. Future comparative studies could be carried out using cross-country samples. Second, due to data availability constraints, we measured digital–intelligent transformation and executive green cognition based on annual report texts and adopted proxy indicators for partial mechanism variables. In future research, these measures could be further validated by incorporating refined data such as ESG reports, executive backgrounds, and actual digital investment figures. Third, we examined three mechanisms through which digital–intelligent transformation affects green innovation from the perspective of internal capability enhancement. However, we did not further investigate whether more complex chain structures exist among these mechanisms. In future research, chain mediation models, structural equation modeling, or more fine-grained data could be used to further examine potential chain relationships. Fourth, although we adopted fixed effect control and multiple robustness tests to alleviate endogeneity bias, unobservable omitted factors may still interfere with empirical results and affect the accuracy of estimation outcomes. In future research, more advanced causal identification strategies could be adopted to further alleviate endogeneity bias.