Green Innovation Driven by Digital Transformation: An Innovation Chain Perspective

Abstract

Taking the innovation chain (IC) as the perspective, we discuss the effect of digital transformation (DT) on enterprises’ green innovation (GI) using data from Chinese listed companies from 2013 to 2021. The results show that DT has a positive effect on enterprises’ GI, and this effect is not only reflected in the quantity of green patent applications but also the GI efficiency and quality. Heterogeneity analysis shows that DT has a positive effect on GI for both large firms and small and medium-sized enterprises (SMEs), but the effect is greater for SMEs. Relative to enterprises that have received governmental incentive-based industrial policies, the effect is smaller in the enterprises that have no incentive-based industrial policies and are subjected to command-based environmental regulations. As the level of industry competition increases from low to high, the impact of DT on GI will grow. However, when the level of industry competition becomes excessively high, the impact will decrease. DT positively affects enterprises’ GI behaviors by facilitating the cultivation of human capital, improving the allocation of innovation resources, and increasing the level of cooperative green innovation.

1. Introduction

Climate change has always been a topic of significant concern in the global economic growth and social development processes [1]. Due to a long-standing focus on heavy industries, a coal-heavy energy structure, and the tightening of global supply chains caused by climate change, achieving high-quality development remains an urgent issue in China. Green innovation (GI) offers a viable solution to address the seemingly contradictory challenges of economic growth and environmental protection. GI refers to new products, processes, technologies, or methods that reduce negative environmental impacts [2,3]. It is often considered to be a model of innovation involving improved products, processes, or management with the goal of achieving environmental sustainability. So, GI is not only a critical source for businesses to shape sustainable competitiveness [4] but also an effective pathway to enhance sustainability and promote green growth [5,6].

The profound integration of digital technology with traditional industries has significantly altered the conventional business model. The comprehensive application of digital technology innovation and artificial intelligence in business activities affects enterprises’ organizational structure, strategic management, production, marketing methods, and operating costs [7]. With the support of modern information technology, the low-carbon industrial production model is becoming the mainstream. Among these, enterprises serve as the central players in industrial digitalization and are also key drivers for GI [8,9,10]. So, can the digital transformation (DT) of business processes and management activities promote GT? If the answer is yes, then it is particularly important to investigate the economic activities through which DT affects GI from an economic and social perspective. As long as we deeply analyze the role of DT in strategic management and economic decision making, we can provide good management enlightenment for enterprises to realize green innovation.

From the perspective of the value chain, some studies put forward that enterprises’ DT will, through micro-mechanisms such as reducing control costs, facilitating the flow and sharing of knowledge, optimizing human capital structures, increasing research and development investment, and promoting business model innovation, affect both core value activities and support value activities, thereby promoting improvements and enhancements in GI [11,12,13,14,15]. From the perspective of industrial organization, some studies suggested that DT reduces transaction costs, promoting the identification of cross-boundary collaborators, and consequently, enhancing the level of GI [16]. DT also generates spillover effects by applying the technological and knowledge spillovers brought about by industrial digitization to upgrade industrial structures, stimulate innovation at the intersection of big data and green technology, and promote the development of green, low-carbon industries [13]. Research from the perspective of macroeconomic growth focuses on analyzing the relationship between the development of digital technology and the quality of economic growth. In the literature, digital technology is regarded as a driver of green development [17,18].

From the perspective of firm size, the impact of DT is considered not only to exist in large enterprises, such as listed companies [12], but also to have a positive effect on the GI performance of small and medium-sized enterprises (SMEs). For instance, Le et al. [15] found a positive relationship between DT and sustainable business performance in SMEs in Vietnam’s food and beverage industry. Additionally, when observing different industries, Huan et al. [14] explored the impact of DT on GI in listed companies in the new energy vehicle industry, discovering a positive effect. Rosa et al. [19] also suggested a positive impact on the food service industry. Le et al. [15] provided further evidence of a positive correlation in the food and beverage industry. Similarly, Rowan et al. [20] found that in the fishing industry, DT positively influences GI. However, Yuan et al. [21] identified a non-linear relationship between DT and GI within agriculture-related firms.

From the perspective of research methods, the published literature [22,23] both used machine learning to discuss the relationship between DT and GI. However, their findings differ. Shao et al. [22] argued that there is a positive relationship between the two concepts, while Han et al. [23] found that the positive correlation between DT and GI exists only in state-owned enterprises and competitive industries. Fan et al. [24] examined the relationship between DT and GI in the supply chain by constructing a tripartite game model involving the government, enterprises, and consumers in green supply chain transformation. Li et al. [25] combined game theory methods with a fixed effects model to discuss the connection between DT and GI.

As evidenced by the literature reviewed, the impact of DT on GI has garnered significant scholarly attention. However, the effects and mechanisms remain unclear, and conclusions from different perspectives are often inconsistent. By integrating theoretical analysis with empirical data from Chinese listed companies, we seek to unveil the ‘black box’ of the process linking enterprise DT to GI from the perspective of the innovation chain (IC). Our study has three objectives:

(1)

To discuss the concept of the IC and define the green innovation chain (GIC);

(2)

To evaluate the impact of DT on GI from the perspective of the GIC;

(3)

To explore the mechanisms through which DT affects GI from the perspective of the GIC.

This study contributes in several ways: (1) Existing studies on the GI effects of DT primarily examine causal logic from the perspective of enterprise value chains, industry chains, and supply chains, with limited research from the IC perspective. However, the IC is crucial for analyzing the process from basic research to the industrialization of high-tech achievements, aiding in understanding the deep-seated logic driving innovation-driven development. This study analyzes the GI effects and mechanisms of enterprise DT from the IC perspective, contributing to the literature on green transformation driven by DT. (2) Research on how enterprise DT affects GI has not reached a consensus, which may be influenced by various factors related to enterprise heterogeneity both internally and externally. Previous studies have mainly examined heterogeneity from regional, industrial, and enterprise-size perspectives. This study, based on the macro–meso–micro support system of the GIC, discusses the variations in the effect of enterprise DT on GI under different macro-industrial policies, market environments, and reputation pressures, thus enriching the study of heterogeneity in the impact of DT on GI. (3) Based on the resource-based view, some studies suggest that enterprise resources are limited, and GI is a high-risk and uncertain innovation activity, with a substitution relationship between independent and cooperative innovation [26]. Thus, the existence and logical operation of cooperative innovation require further theoretical clarification. This study, through theoretical and empirical analysis, discusses the mechanism of cooperative innovation from the IC perspective, enriches the understanding of the mechanism, and provides evidence for improving GIC efficiency.

2. Literature Review and Research Hypothesis

2.1. Definition and Research Progress of Innovation Chain

The IC is defined from a chain perspective for defining innovative behavior, indicating a chain-like process of input to output, from basic research to the production and commercialization of innovative results. Unlike concepts such as the innovation network or innovation ecosystem, the IC stems from an academic interest in the innovation process.

Freeman [27] expanded on the Schumpeterian concept of innovation, dividing innovation into two processes: invention and diffusion. As the path of technological progress continued to evolve, the boundaries between knowledge innovation and technological innovation increasingly merged, further extending innovation into four stages: basic research, applied research, development, and post-research (also known as diffusion) [28]. In the context of the global technological revolution achieving disruptive breakthroughs, the basic research stage, namely knowledge innovation, has become increasingly critical in innovation competition.

Early research on the IC primarily focused on exploring its content, including defining the IC from different perspectives such as knowledge innovation, technological innovation, and changes in the subjects of innovation. Researchers also investigated its characteristics [29]. Regardless of the perspective taken, the core concepts include (1) several functional nodes of the IC, involving various types of actors in innovation activities such as enterprises, governments, financial institutions, research institutions, intermediaries, and more; (2) relationships between these functional nodes, manifesting as cooperation, strategic alliances, and other specific forms; (3) the sharing of information and knowledge among these nodes, including formal and informal channels and mechanisms; and (4) the external support system for the operation of the IC, encompassing institutions, macro-industrial policies, markets, changes in consumer demand, and other macro–micro factors. Another body of literature interpreted the models and structures of the IC from perspectives like enterprise value creation [30], innovation boundaries [31], and global value chains [32]. These studies also analyzed and discussed various heterogeneous influencing factors on different dimensions to understand and analyze the key links, influencing factors, and interactions in the innovation process. The recent literature has focused on the inherent logic and interactive paths of the IC with the industrial chain [33] and the value chain [34].

As the scope and extension of the IC continue to expand, its importance has risen to the level of national strategy. In China’s 14th Five-Year Plan, enhancing the overall efficiency of the IC is identified as a crucial aspect of driving innovation-led development. Improving the connectivity between different stages and functional nodes of the IC and optimizing the integration between the IC and the industrial chain and value chain holds great significance for high-quality economic development driven by innovation. The key to achieving these goals lies in breakthroughs and development in general technology, disruptive technology, and cutting-edge technology.

2.2. The Theoretical Logic of Enterprise Digital Transformation Affecting Green Innovation

According to IC, the GIC can be divided into three stages: basic research, applied research, and the adoption of new technologies. Basic research involves knowledge innovation, with universities and research institutes as the main actors. This type of innovation may precede the green technology needs of enterprises and has the characteristics of a public good. Applied research involves converting knowledge into new technologies based on the green technology needs of enterprises. This is a critical link in the GIC, as it transforms knowledge into innovation that can be commercialized. The adoption of new technologies includes innovations in green business models and market innovations, which represent the stage where GI outcomes are transformed. From the value sources of GIC’s three stages, the value of basic research comes from the knowledge innovation of universities, scientific research institutions, and other subjects. The value of applied research comes from the green technology innovation demand of enterprises, which promotes the transformation of knowledge innovation into green technology innovation. The value created in the adoption stage from the application of new technology to meet the needs of consumers. But the total value of GIC is not a simple sum of the three stages. Under the interaction of different subjects in three stages, the total value of GIC is formed by nonlinear summation.

Enterprises are important functional nodes in the GIC. The impact of DT on GI can be seen in two main aspects. Firstly, it is reflected in the optimization of the internal functional nodes of the GIC. Secondly, it is manifested in the facilitation of the connections between different functional nodes, therefore enhancing the overall efficiency of the GIC and increasing its total value (Figure 1).

Regarding the different stages of the GIC, DT can be seen as an effort to apply digital technology to the goals of applied research. It can also be seen as the enterprises’ efforts in the adoption of digital technology. By undergoing DT, enterprises gain access to digital technologies with the characteristics of generality, openness, and availability. These technologies serve as the foundation for the fusion of digital and green technologies and the incubation of new technologies. For example, digital twin technology provides a virtual space for testing the feasibility of new technologies, reducing the risk and cost of research and development. Additionally, big data analytics enhances the ability to analyze vast and diverse structured or unstructured data sources, facilitating the discovery of potential issues in existing GI systems and optimizing the GI process. Some artificial intelligence technologies can replace human labor in high-risk, high-precision experiments, expanding the scenarios of digital technology application in green products. Digital technology also enables enterprises to connect with consumers and innovate in green business models. Companies can better understand consumer demands for green products and services, optimize innovation strategies, and adjust innovation resource allocation after DT. Digital technology has the potential to transform the direction of technology demand, leading to a shift from pollution-oriented to green-oriented innovation. Given these considerations, the hypotheses are put forward.

Hypothesis 1 (H1): 

Enterprise DT has a positive effect on GI.

While digital technology possesses generality, it also comes with complexity and skill biases. The success of the fusion of digital and green technologies, as well as the conversion of digital technology into secondary innovation, relies heavily on the human capital within enterprises. On one hand, enterprises need employees with knowledge and skills related to digital technology to facilitate its diffusion and application. This drives some enterprises to actively recruit employees who are proficient in digital technology during the DT process to optimize their human capital structure for GI. On the other hand, one of the channels for the formation of human capital is through employee learning and interaction, promoting the diffusion of heterogeneous knowledge. Digital management tools, by optimizing information flows and standardizing communication processes, reduce internal control and communication costs and encourage active learning and idea exchange among employees, contributing to the formation of new ideas and knowledge for GI.

Moreover, the application of digital technology will change the way companies train their employees, providing various forms of training such as remote training, self-directed learning, and personalized training. This diversification of training methods enhances the development of human capital through multiple channels. Employee training not only promotes the cultivation of human capital within the enterprise but also provides positive feedback for digital technology adoption, creating a positive loop for DT. The positive relationship between human capital and GI has been certificated in previous research [35,36].

Certainly, the realization of GI value by enterprises depends on the interaction of other resources such as funding, technology, and information. The new factor of data not only replaces traditional high-energy, low-output factors but also affects the direction of technological demand, subsequently influencing GI. Data-driven decision-making models facilitate quick responses to market demands and optimization of innovation strategies, improving the efficiency of GI supply and demand matching. DT improves GI by enhancing the optimization of human capital structure and improving the allocation of innovation resources. This is reflected in the GIC as an increase in the value of individual functional nodes. Therefore, the following two research hypotheses are proposed.

Hypothesis 2 (H2): 

Enterprise DT promotes GI through human capital cultivation.

Hypothesis 3 (H3): 

Enterprise DT promotes GI by improving the ability to allocate innovative resources.

Furthermore, we expand our analytical perspective to the interactive dimension of various functional nodes within the IC. In the context of conventional technological advancement pathways, a majority of enterprises’ GI activities are typically confined within the boundaries of the enterprise itself, and the realization of value from GI remains limited to the stages of applied research or the adoption of new technologies. The compartmentalization of innovation activities results in fragmentation within the GIC, particularly between the phases of basic research and subsequent stages, rendering innovation resources challenging to effectively allocate along the upstream and downstream of the IC, thereby impacting the overall value of GI. According to research conducted in the published literature [37], innovation activities concentrated solely on individual functional nodes within the IC, i.e., independent organizational innovation, especially within enterprises with weak innovation capabilities, constitute a major contributing factor to the decline in the quality of green technological innovation. Moreover, compared to general innovation, GI is characterized by a higher level of asset specificity and uncertainty. Cooperative innovation represents a significant form of resource allocation within the IC. In practice, cooperative innovation with supply chain partners, competitors, and even academic and research institutions has become a crucial strategic approach for many enterprises to reduce uncertainty and enhance the value of GI [38,39].

Although DT is mainly an initiative taken by enterprises, it can promote cooperative innovation and improve the efficiency of GIC. On the one hand, DT enhances the efficiency of interacting with other stakeholders in the GI process. On the other hand, digital technology has significantly improved information asymmetry and reduced the transaction costs of innovation participants. When the frequency of cooperative innovation between enterprises, universities, and research institutions increases, it will not only bring more external resources to enterprises but also stimulate GI with more technological novelty, thus improving the quality of GI [40]. Based on the above analysis, the following hypothesis is proposed:

Hypothesis 4 (H4): 

Enterprise DT promotes GI by enhancing cooperative innovation capabilities.

Figure 1 presents our research framework, which illustrates the relationship between DT and GI. This includes both the direct effects of DT on GI and the indirect effects through three channels: human capital cultivation, innovation resource allocation, and collaborative innovation. Additionally, Figure 1 outlines the hypotheses of our study.

3. Methodology

3.1. Research Sample and Data Source

In this study, companies listed in the Shanghai and Shenzhen A-share markets from 2013 to 2021 were selected as the sample. The following data-processing steps were applied to the original data: (1) excluding companies in the financial industry; (2) excluding Special Treatment (ST), * Special Treatment (*ST), and Particular Transfer (PT) companies at year-end, as well as companies within one year of their IPO; (3) excluding samples with missing key variables in the baseline regression; and (4) truncating continuous variables at the 1% level to reduce the impact of outliers. After these steps, there are a total of 16,366 observations.

3.2. Variable Setting

Green innovation (GI). GI has been a focal point in various academic disciplines such as environmental economics, innovation economics, strategic management, and industrial organization. Following the perspectives of scholars like the authors of [37,41], patent application data not only reflect the output of enterprises’ GI activities but can also provide insights into the structure and areas of GI based on patent categories. Therefore, this study uses the number of green invention patent applications to measure enterprises’ GI activities.

Digital transformation (DT). According to the views of scholars such as those in the published literature [42,43], DT relies on the development of new-generation digital technologies (such as blockchain, artificial intelligence, and big data analytics), making digital technology the core of transformation. Building on this understanding, this study, following the approach of Wu et al. [44], measures the level of corporate DT based on the frequency of the term “digital transformation” in annual reports of listed companies. The data mainly come from the CSMAR database, and due to their right-skewed distribution, they are naturally log-transformed by adding 1.

Controls (CV). Based on the methodology proposed in the analysis of the digital transformation and technological innovation of office business in the published literature [12,45,46,47], this study controls for a series of variables that may affect GI behavior, including enterprise age (AGE), ownership (SOE), enterprise size (SIZE), the duality of roles (GOV), leverage ratio (LEV), profitability (ROA), business revenue (BR), Tobin Q (TQ) value, management ownership (MO), and the Herfindahl–Hirschman index (HHI) (a composite index used to measure industry concentration) for industry concentration. Descriptive statistics of the main variables are provided in

Table 1. Descriptive statistics of variables.

Variables	Symbol	Definition	N	Mean	Sd	Min	Max
Green innovation	GI	Logarithm (base e) of the number of green invention patent applications by manufacturing enterprises, with a constant of 1 added.	16,366	0.761	0.000	3.664	0.377
Digital transformation	DT	Logarithm (base e) of the word frequency related to digital transformation found in the manufacturing enterprises’ annual reports, with a constant of 1 added.	16,366	1.443	0.000	5.209	1.635
Enterprise age	AGE	Logarithm (base e) of the difference between the year of sample observation and the year of the enterprise’s establishment.	16,366	0.324	1.946	3.497	2.871
Enterprise ownership	SOE	Binary variable, where 1 represents state-owned enterprises, and 0 represents private enterprises.	16,366	0.452	0.000	1.000	0.287
Enterprise size	SIZE	Logarithm (base e) of the number of total number of employees.	16,366	1.156	5.384	11.086	7.747
Duality of roles	GOV	Binary variable, where 1 indicates that the chairman and the general manager roles are held concurrently by the same person, and 0 indicates otherwise.	16,366	0.465	0.000	1.000	0.315
Leverage ratio	LEV	Ratio of total liabilities to assets.	16,366	0.190	0.060	0.859	0.401
	ROA	Return on assets, ratio of net income to total asset.	16,366	0.065	−0.276	0.201	0.039
Business revenue	BR	The increase rate of operating revenue.	16,366	0.809	−0.936	5.877	0.207
Tobin Q	TQ	Proportion of market capitalization to total assets.	16,366	1.362	0.858	8.690	2.161
Management ownership	MO	Proportion of all shares held by directors, supervisors and executives.	16,366	0.201	0.000	0.675	0.163
Herfindahl–Hirschman index	HHI	Herfindahl–Hirschman Index	16,366	0.138	0.027	0.862	0.135

.

3.3. Econometric Model

Based on the research objectives and hypotheses, the econometric model is specified:

$${GI}_{i,t}={\beta }_0+{\beta }_1{DT}_{i,t-1}+{\beta }_2{CV}_{i,t}+{\gamma }_j+{\mu }_t+{\epsilon }_{i,j,t}$$

(1)

In Equation (1), the dependent variable is GI, measured using the number of green invention patent applications by companies. As this variable exhibits a noticeable right-skewed distribution, it is log-transformed by adding 1 before entering the regression model. The core explanatory variable is corporate DT (DT), which is lagged by one period in the model. CV represents all the control variables mentioned earlier. ${\gamma }_j$ represents industry-fixed effects, and ${\mu }_t$ represents firm-fixed effects. Since our data cover multiple industries excluding the financial sector, it is necessary to control for industry-fixed effects. Controlling for time-fixed effects is essential to account for factors that affect all individuals uniformly over time. ${\epsilon }_{i,j,t}$ represents the robust standard errors clustered at the firm level. ${\beta }_0$ is a constant term. ${\beta }_1$ and ${\beta }_2$ are the regression coefficients waiting to be calculated.

4. Results

4.1. Baseline Regression

We estimate Equation (1) by using the ordinary least square (OLS). Columns (1) and (2) in

Table 2. The result of baseline regression.

Variables	(1)	(2)	(3)	(4)
DT	0.097 ***	0.073 ***	0.379 ***	0.218 ***
	(0.012)	(0.011)	(0.074)	(0.072)
Constant	0.236 ***	−0.882 ***	0.742 ***	−6.857 ***
	(0.020)	(0.181)	(0.202)	(1.206)
CV	No	Yes	No	Yes
N	12,837	12,837	12,646	12,646
Industry FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Adj R2	0.128	0.194
Pseudo R2			0.247	0.492

Note: robust standard errors clustered at the firm level in parentheses; *** p < 0.01. The observed values in columns (3) and (4) decrease because the ReLU method of the Poisson Pseudomaximum Likelihood automatically dropped the separated observations.

present the results estimated using a panel-fixed effects (FE) model. Both columns (1) and (2) show that, after controlling for industry and year-fixed effects, the core explanatory variable, corporate DT, has a positive effect on GI. Although the coefficient decreases slightly after including control variables (β = 0.073), it remains statistically significant. Given that patent data are typically counted, taking a direct logarithm might result in biased estimates, and because there are many zero values in green patent data, it is challenging to determine whether these zeros indicate a lack of GI or a strategic disclosure of R&D activities by companies. The results of the Poisson Pseudomaximum Likelihood are presented in columns (3) and (4). The results show that DT still has a positive effect on GI. Column (4) suggests that for every one-unit increase in corporate DT, the average number of GI will increase by exp (0.218) ≈ 1.2435 times. This can be interpreted as the incidence rate ratio (IRR), indicating that the GI occurrence rate for transformed companies is 24.35% higher compared to non-transformed companies. The results of the baseline regression show that companies that have experienced DT are more active in GI than those that have not experienced DT, which is consistent with the existing literature in this field [12]. According to the results of the fixed-effect model and Poisson regression models in Table 2, it can be seen that DT has a significant positive impact on GI, which indicates that H1 has been proved.

Furthermore, as DT relies on different technologies, we assess the effect of DT on GI from five technology dimensions: Artificial Intelligence (AI), Blockchain Technology (BLOCKCHAIN), Cloud Computing (CLOUD), Big Data Analytics (BDA), and Digital Technology Application (DIGAP). The results correspond to columns (1), (2), (3), (4), and (5) in

Table 3. The difference between different digital technologies.

Variables	(1)	(2)	(3)	(4)	(5)
L.AI	0.108 ***
	(0.018)
L. BLOCKCHAIN		0.179
		(0.133)
L. CLOUD			0.100 ***
			(0.017)
L. BDA				0.068 ***
				(0.017)
L. DTAP					0.063 ***
					(0.012)
Control variable	Yes	Yes	Yes	Yes	Yes
Constant	−0.860 ***	−0.861 ***	−0.864 ***	−0.856 ***	−0.863 ***
	(0.180)	(0.182)	(0.180)	(0.181)	(0.182)
N	12,837	12,837	12,837	12,837	12,837
Industry FE	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes
Adj R2	0.192	0.185	0.194	0.188	0.190

Note: robust standard errors clustered at the firm level in parentheses; *** p < 0.01.

. According to Table 3, all five technology dimensions have a positive effect on GI. However, the coefficient for BLOCKCHAIN is not statistically significant. This may be due to the limited adoption of Blockchain Technology among Chinese listed companies, especially smaller enterprises, which may require a significant fixed-cost investment. It could also be due to a lag in the impact of Blockchain Technology on GI [48]. At present, some companies’ DT is still at the stage of optimizing production efficiency and improving organizational processes using general technologies like ICT, Big Data, and ERP systems, and the transformation driven by Blockchain Technology in the context of green business models is ongoing.

4.2. Robustness Tests

4.2.1. Endogeneity Tests

Endogenous problems are also a concern in this study. On one hand, achieving GI requires companies to adopt more environmentally friendly and efficient technologies and production methods, which will lead to improvements and optimizations in production and business processes, increasing the demand for DT. On the other hand, GI is also a form of innovation capability for companies. In the process of promoting GI, companies accumulate a significant amount of knowledge, technology, and management experience. This experience can help companies achieve DT more effectively. Research has already shown a positive correlation between DT and a company’s knowledge accumulation, management, and decision-making experience [48].

To mitigate endogenous problem, in the baseline regression, we lagged the DT by one period, reducing some degree of endogeneity. Next, we used the spherical distance between the registered city of listed companies and Hangzhou as an instrumental variable. The rationality of choosing this instrumental variable lies in the fact that Hangzhou is the birthplace of digital finance in China. Digital finance is an integral part of urban digital economic development and is closely related to the degree and motivation of corporate DT, meeting the relevance requirement of instrumental variables. Additionally, the spherical distance between the registered city of listed companies and Hangzhou has a physical attribute and is fixed, with minimal direct impact on a company’s GI behavior, meeting the exogeneity requirement. For the estimation using instrumental variables, we used two-stage least squares (2SLS) as the regression method. The results are presented in column (1) of

Table 4. Results of endogeneity test.

Variables	(1)	(2)	(3)	(4)
	2SLS	DID	OLS	OLS
DT	0.047 **		0.069 ***	0.064 ***
	(0.019)		(0.011)	(0.011)
Digdum × Yeardum		0.053 **
		(0.021)
CV	Yes	Yes	Yes	Yes
N	12,837	12,837	12,837	12,837
Industry FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
City FE	No	No	No	Yes
F statistics	1359.55
Kleibergen–Paap rk LM statistic	482.643
Cragg–Donald Wald F statistic	8966.931
Adj R2	0.087	0.185	0.200	0.231

Note: robust standard errors clustered at the firm level in parentheses; *** p < 0.01, ** p < 0.05.

. Column (1) in Table 4 shows that the F-statistic of the first-stage regression in the 2SLS model is 1359.55, indicating that the instrumental variable meets the relevance requirement. In the second-stage regression, the coefficient of DT is significantly positive at the 5% statistical level. The Kleibergen–Paap rk LM statistic is 482.643, suggesting that the instrumental variable passed the identification test. The Cragg–Donald Wald F statistic is 8966.931, indicating that the instrumental variable passed the weak instrumental test. This suggests that, after addressing endogeneity with instrumental variables, the positive impact of DT on GI remains robust.

We also used the first introduction year of China’s Digital China strategy as an exogenous shock, removing interference from other factors at the enterprise level. The Digital China strategy was first proposed in 2015, and based on this, we designed a year fixed effect variable Yeardum. When the sample year is before the policy introduction, the Yeardum is set to 0, and when it is after the policy introduction, the Yeardum is set to 1. Furthermore, we generated a dummy variable for DT (Digdum) based on the company’s annual report. Based on these settings, we created an interaction term Digdum × Yeardum. To assess the policy effect of the Digital China strategy as an exogenous shock, we established a difference-in-differences (DID) model, as shown in Equation (2).

$${GI}_{i,t}+{\beta }_0+{\beta }_1{Digdum}_i\times {Yeardum}_t+{\beta }_2{CV}_{i,t}+{\gamma }_j+{\mu }_t+{\epsilon }_{i,j,t}$$

(2)

In Equation (2), GI represents green innovation, and Digdum is a dummy variable indicating whether a company has adopted DT based on its annual report; it is assigned a value of 1 if DT-related terms are present, otherwise it is 0. The Yeardum is a time dummy variable, with a value of 0 for years before 2015 and 1 for 2015 and subsequent years. CV represents control variables with the same meaning as in Equation (1), ${\gamma }_j$ is industry-fixed effects, and ${\mu }_t$ is year-fixed effects. ${\epsilon }_{i,j,t}$ represents robust standard errors clustered at the firm level. In Equation (2), the focus is on the regression coefficient of the interaction term Digdum × Yeardum. The results using the DID method are presented in column (2) in Table 4. Column (2) indicates that the estimated coefficient of the interaction term (Digdum × Yeardum) is statistically significant at the 5% level, suggesting the robustness of the results.

In addition, endogeneity in this study could also be influenced by omitted variables. In the baseline regression, we found that when all control variables were added, the regression coefficient decreased, and the estimates were absorbed by unobservable factors. To reduce interference from omitted variables, we included control variables at the city level that may affect GI. These include per capita GDP in the city (CITYGDP), the proportion of secondary industry value added to GDP (CITYIND), the level of foreign direct investment in the city (CITYFDI), the balance of loans and deposits in financial institutions at the end of the year (CITYFINANCE), and the level of digital economic development in the city (CITYDIG). Consistent with the baseline regression, the OLS is used here. The results are shown in columns (3) and (4) of Table 4. Columns (3) and (4) demonstrate that, after adding city-level control variables, the Adj R² of the model improves, and the coefficient for DT is statistically significant at the 1% level. This suggests that the impact of corporate DT on GI remains robust after addressing endogeneity.

4.2.2. Robustness Tests in Other Ways

(1)

Alternative Measurement of Independent Variables. Firstly, we added an alternative indicator for DT, using the ratio of intangible assets related to digital technologies (DT_assets) as a robustness check. According to Tao et al. [49], we identified intangible assets related to digital technologies from the total intangible assets of companies and then calculated the ratio of these intangible assets relative to total assets as an alternative indicator for DT. The result is presented in column (1) of

Table 5. Results of robustness tests.

Variables	(1)	(2)	(3)	(4)	(5)
DT_assets	0.646 ***
	(0.124)
DT_FCA		0.303 ***
		(0.049)
DT			0.025 ***	0.098 ***	0.084 ***
			(0.008)	(0.015)	(0.020)
Constant	−0.883 ***		0.040	−1.345 ***	−1.396 ***
	(0.182)		(0.503)	(0.223)	(0.354)
N	12,837	12,837	12,837	8438	4766
CV	Yes	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes
Enterprise FE	No	No	Yes	No	No
Adj R2	0.192	0.195	0.687	0.217	0.219

Note: robust standard errors clustered at the firm level in parentheses; *** p < 0.01. The decrease in the observations in column (4) is due to the exclusion of samples that did not disclose DT vocabulary in the annual report. The observations in column (5) decrease because of the use of a balance panel.

. From column (1) of Table 5, it can be seen that the coefficient of DT_assets is significantly positive at the statistical level of 1%. Secondly, we conducted a principal component analysis (PCA) of the five technology dimensions to remove redundant information provided by individual indicators and generate a new DT indicator (DT_FCA). The results are shown in column (2) of Table 5. The results in column (2) indicate the robustness of the results.

(2)

Alternative Estimation Methods. In the fixed-effects model designed in the baseline regression, only industry and year-fixed effects were controlled, which might overlook some individual-level factors affecting companies. Therefore, we added controls for individual company effects to the model. The results are presented in column (3) of Table 5. The results in column (3) of Table 5 indicate that even after controlling for individual company effects, DT is statistically significant at the 1% level.

(3)

Eliminate the influence of the strategic behavior of exaggerating DT in the annual report. Since the DT indicator used in this study is obtained from annual reports, companies may engage in exaggeration in their disclosures, resulting in an overestimation of the degree of DT. We also observed that there are observations with zero DT in the sample, indicating that some companies may not have undergone DT. Therefore, we re-estimated the regression after removing samples with zero DT. The results are shown in column (4) of Table 5. In column (4) of Table 5, the coefficient of DT is 0.098 and statistically significant at the 1% level.

(4)

Consider the entry and exit of enterprises. Entry and exit behaviors of companies are essential factors influencing industry competition, which, in turn, affect a company’s GI [50]. To account for this, we used a balanced panel format that excludes the impact of company entry and exit. The results are shown in column (5) of Table 5. The entry and exit behavior of companies, as shown in the regression results in column (5) of Table 5, still supports the conclusion of the baseline regression.

5. Heterogeneity Analysis

5.1. Heterogeneity in Firm Size

Does the relationship between DT and GI vary with firm size? Liu et al. [12] argued that the positive impact of DT is not only evident in large firms but also in SMEs. Le et al. [15] found a positive relationship between DT and GI among SMEs in Vietnam’s food and beverage industry. Although the impact is positive in large firms and SMEs, the degree of this impact may differ between them. This difference arises because large firms, compared to SMEs, often have more extensive social networks in the market, which facilitates easier connections with other entities involved in GIC, such as universities and research institutions. Additionally, DT requires a fixed-cost investment, and in this regard, large firms have a financial advantage over SMEs. On the other hand, large firms typically have more complex organizational structures than SMEs, which can lead to relative disadvantages in terms of information transmission and resource allocation efficiency. Therefore, the impact of DT on GI may vary depending on the size of the firm. In this study, firm size is proxied by revenue, with sample firms classified into SMEs and large firms. Specifically, firms are grouped by the quartile of revenue, where the first and second quartiles represent SMEs, and the third quartile represents large firms. The results are presented in

Table 6. Heterogeneity analysis results of firm size.

Variables	(1)	(2)
	Small and Medium Firms	Large Firms
DT	0.055 ***	0.035 *
	(0.010)	(0.018)
Constant	0.031	0.103
	(0.145)	(1.219)
N	8188	4649
CV	Yes	Yes
Industry FE	Yes	Yes
Year FE	Yes	Yes
Adj R2	0.136	0.739
The empirical p-value of Fisher’s permutation test	0.002

Note: robust standard errors clustered at the firm level in parentheses; *** p < 0.01, * p < 0.1. The empirical p-value is the result of conducting a Fisher’s permutation test on the coefficients of the two groups. In all subsequent tables requiring grouped tests, we provide the empirical p-value.

.

The results in Table 6 indicate that in SMEs, the coefficient of DT is 0.055, which is significant at the 1% level. In contrast, in large firms, the coefficient of DT is 0.035, which is significant at the 10% level. Fisher’s permutation test was conducted on the DT coefficients from the grouped regressions, yielding an empirical p-value of 0.002, suggesting a significant difference in the DT coefficients between large firms and SMEs. This shows that while DT has a promoting effect on GI for both large firms and SMEs, the effect is stronger in SMEs. This is because SMEs typically have a flatter organizational structure, which enhances the efficiency of innovation resource allocation and allows for a more agile response to environmental changes.

5.2. Heterogeneity in Industry Policies

New technology adoption exhibits strong path-dependence characteristics [51], and the same holds true for digital technology. While the rapid development of digital technology has lowered the hardware and software costs, thereby creating conditions for enterprise DT, the underlying technological investments in digitalization still entail relatively high fixed costs and opportunity costs. Liu et al. [52] estimated the investment threshold for enterprise DT to be between CNY 1 million and 2 million. The existence of fixed costs may impact enterprises’ willingness to undergo DT, and when the level of DT is relatively low, its natural influence on GI also decreases. Industrial policies, through incentives, restrictions, or phasing out, can effectively guide enterprises’ production, investment, and other behaviors in the short term. Different types of industrial policies have varying impacts on enterprise DT behavior, and these differences are transmitted to the process through which DT affects GI. This section explores the differential impacts of incentive-based and restriction-based industrial policies.

R&D subsidies are a common tool of incentive-based industrial policies. To assess the heterogeneity of incentive-based industrial policies, we first conducted searches using keywords such as “R&D”, “research”, “development”, “innovation”, “creation”, “technology”, “strategic emerging industries”, “high-tech”, “industry-university-research”, and similar terms to retrieve data from the government subsidy details in the financial statements of listed companies, which we then aggregated into annual data. Subsequently, we assigned a value of 1 to samples with R&D subsidy amounts greater than 0 and a value of 0 to samples with amounts less than or equal to 0, grouping them for regression analysis.

The grouped regression results for R&D subsidies are shown in columns (1) and (2) in

Table 7. Results of heterogeneity tests.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Non-Subsidy	Subsidy	Non-Pollution	Pollution	Low-HHI	Medium-HHI	High-HHI
DT	0.066 ***	0.094 ***	0.085 ***	0.033 *	0.043 **	0.118 ***	0.048 ***
	(0.011)	(0.027)	(0.013)	(0.019)	(0.018)	(0.018)	(0.016)
Constant	−0.864 ***	−1.144 ***	−0.818 ***	−1.189 ***	−1.096 ***	−0.800 **	−0.893 ***
	(0.193)	(0.326)	(0.190)	(0.318)	(0.300)	(0.311)	(0.295)
N	10,088	2749	8714	4123	4254	4340	4243
CV	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Adj R2	0.195	0.191	0.194	0.213	0.153	0.197	0.235
The empirical p-value of Fisher’s permutation test	0.016		0.000

Note: robust standard errors clustered at the firm level in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

. From columns (1) and (2), it is apparent that while DT has a positive effect on GI, regardless of whether there are R&D subsidies, this effect is greater for companies receiving R&D subsidies. This is because DT necessitates substantial allocation of resources by enterprises, while R&D subsidies can effectively compensate for the costs when undergoing DT. R&D subsidies also contribute to enhancing enterprises’ capacity to bear risks [53], subsequently transmitting this positive influence to the process through which DT affects GI. The cost compensation effect of R&D subsidies further stimulates enterprises to innovate in digital technology, promotes the diffusion of new technology in the field of green technology, and enhances the level of GI.

Next, we examine the effect of restriction-based industrial policies. Command environmental regulations are commonly used environmental management measures and can be seen as a tool of restriction-based industrial policies. To standardize and strengthen environmental information disclosure by listed companies and promote corporate environmental protection responsibilities, the Chinese Ministry of Environmental Protection issued the “Catalog of Classification and Management of Environmental Protection Inspections for Listed Companies” in June 2008, which identified 14 industries, including thermal power and steel, as heavily polluting industries. Compared to other companies, companies in heavily polluting industries have a greater impact on the environment due to their production activities and, therefore, face greater regulatory pressures. To examine the impact of command environmental regulations, we assigned a value of 1 to sample companies in heavily polluting industries and 0 otherwise.

The results are shown in columns (3) and (4) in Table 7. From columns (3) and (4), it can be observed that DT has a positive effect on GI, whether or not the companies belong to heavily polluting industries, but this effect is greater for samples in non-heavily polluting industries. This can be explained by the fact that enterprises’ GI activities incur higher costs and uncertainty. Under the context of command environmental regulations, enterprises also need to bear certain costs to fulfill their environmental protection responsibilities. However, simultaneously, enterprise DT also requires the investment of fixed costs, especially in the early stages of transformation, where the initial fixed costs are relatively high, and the payback period is relatively long. Under the influence of several overlapping costs, the GI effects obtained by enterprises through DT will be diminished.

5.3. Heterogeneity in Market Competition Level

In addition to industrial policies, the level of industry competition is another significant factor influencing DT [50]. When enterprises operate in industries with low levels of competition, new entrants face higher barriers to investing in digital technology with the goal of enhancing their GI capabilities. Apart from resistance from incumbent enterprises, banks and investors are more inclined to invest in enterprises that have already demonstrated successful DT. According to Bai et al. [54], the level of market competition affects a company’s agency costs. In highly competitive markets, managers are more likely to consider customer opinions to maintain a competitive edge. Based on this perspective, an increase in market competition in the digital economy will prompt managers to adopt digital technologies to enhance the company’s competitiveness. Indeed, especially under the impact of COVID-19, the adoption of digital technologies for DT has become crucial for coping with intense market competition. As Jaskeviciute et al. [55] point out, COVID-19 has presented organizations with new challenges, requiring them to adapt to maintain vitality, ensure survival, and promptly respond to and adjust their performance and workforce management strategies. Consequently, enterprises face a certain degree of financial constraints. As industry competition intensifies, it becomes easier for enterprises to introduce green products into the market, gaining cost compensation for both DT and GI. To test whether the impact of DT on GI varies across different levels of industry competition, we divided the HHI into three groups based on percentiles, namely Low-HHI, Medium-HHI, and High-HHI, representing low to high levels of competition. The results are shown in columns (5) through (7) of Table 7.

From columns (5) and (6) of Table 7, it can be observed that as market competition increases, enterprises’ motivation to gain market share and competitive advantages in green products through DT strengthens. Additionally, in highly competitive industries, enterprises’ flexibility in selecting green cooperative innovation partners also increases [56], providing more options for choosing green technological innovation modes. However, the impact of industry competition is not linear. Column (7) of Table 7 shows that in highly competitive industries, the coefficient of DT decreases compared to that in column (6). When industry competition becomes excessively high, enterprises’ profit margins may be compressed [57]. In such cases, the focus of DT may shift towards cost reduction, expanding the scale of existing products, and other purposes, leading to a reduction in R&D investments in GI.

5.4. Heterogeneity in Social Environmental Concern

Apart from government macroeconomic industrial policies, the level of societal environmental concern is another crucial factor influencing the shift towards GI. On one hand, an increase in societal environmental concern can act as a means to mitigate information asymmetry in the execution of macroeconomic industrial policies. On the other hand, heightened societal environmental concern also reflects the preference for green consumption among residents [58,59], which leads to a transmission of demand for green products along the GIC, thereby providing rationality for green cooperative innovation. Additionally, in response to external pressure from societal environmental concerns, companies may proactively disclose their environmental protection responsibilities to enhance their reputation, gain market competitiveness, and develop new products. This falls under the category of internally driven proactive disclosure behavior. From an internal perspective within the company, such proactive disclosure behavior also reflects societal environmental concerns. Hence, we examine the heterogeneous effects of societal environmental concern from both external pressure and internal information disclosure aspects, namely, the “pressure-reflection” perspective.

Concerning external environmental concern pressure, we draw inspiration from Cheng and Liu [60] and use the Baidu Index for the keyword “environmental pollution” among Chinese internet users as a measurement indicator, sourced from the Baidu Index official website. We divided the sample companies into two groups based on the percentiles of their Baidu Index, namely High-out-attention and Low-out-attention.

The grouped results are shown in columns (1) and (2) of

Table 8. Heterogeneity analysis results of social environmental concern degree.

Variable	(1)	(2)	(3)	(4)
	Low-Out Attention	High-Out Attention	Low-Quality Disclosure	High-Quality Disclosure
DT	0.067 ***	0.079 ***	0.044 **	0.075 ***
	(0.015)	(0.014)	(0.021)	(0.012)
Constant	−0.559 **	−1.157 ***	−0.423	−0.967 ***
	(0.240)	(0.245)	(0.286)	(0.196)
N	6281	6556	1631	11,206
CV	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Adj R2	0.179	0.213	0.101	0.206
The empirical p-value of Fisher’s permutation test	0.170		0.074

Note: robust standard errors clustered at the firm level in parentheses; *** p < 0.01, ** p < 0.05.

. From columns (1) and (2), it can be observed that regardless of whether external environmental concern is high or low, DT has a positive impact on GI behavior. However, the empirical p-value of Fisher’s Permutation test is not significant, indicating that there is no statistically significant difference in the estimated coefficients of DT between the high and low external environmental concern groups.

Regarding internal information disclosure, the Shanghai Stock Exchange and the Shenzhen Stock Exchange assess the annual information disclosure of listed companies, categorizing the results into four levels: A, B, C, and D. This assessment can reflect the quality of a company’s environmental protection responsibility disclosure, thus providing insights into the proactive response strategies of companies towards societal environmental concern. We have reclassified firms with C and D ratings as having low information disclosure quality (Low-quality-disclosure) and those with A and B ratings as having high information disclosure quality (High-quality-disclosure). The grouped regression results are shown in columns (3) and (4) of Table 8.

The grouped results in columns (3) and (4) of Table 8 indicate that proactive disclosure behavior by companies is a positive response to societal environmental concern. When a company’s information disclosure quality is higher, it can stimulate public preference for green products through the power of reputation, thereby gaining new product markets and generating positive feedback for innovation.

In summary, whether it is external environmental concern pressure or a company’s proactive response strategy, both influence the relationship between DT and GI. However, comparatively, the impact of a company’s proactive response strategy on this relationship is stronger. Such proactive response strategies benefit stakeholders and the public in understanding a company’s environmental protection actions. When a positive reputation effect is formed, it can provide cost compensation for DT through new product revenue.

6. Mediation Mechanism Analysis

We have validated the impact of corporate DT on GI behavior and analyzed how this impact varies across various heterogeneous dimensions in previous sections. In this section, based on an intermediate effect model, we examine potential transmission mechanisms. The econometric model is set up as follows:

$${GI}_{i,t}={\beta }_0+{\beta }_1{DT}_{i,t-1}+{\beta }_2{CV}_{i,t}+{\gamma }_j+{\mu }_t+{\epsilon }_{i,j,t}$$

(3)

$${MEDIATOR}_{i,t}={\alpha }_0+{\alpha }_1{DT}_{i,t-1}+{\alpha }_2{CV}_{i,t}+{\gamma }_j+{\mu }_t+{\epsilon }_{i,j,t}$$

(4)

$${GI}_{i,t}={\phi }_0+{\phi }_1{DT}_{i,t-1}+{\phi }_2{{MEDIATOR}_{i,t}+{\phi }_{3}CV}_{i,t}+{\gamma }_j+{\mu }_t+{\epsilon }_{i,j,t}$$

(5)

In Equation (4), the independent variable is the lagged one-year corporate DT, and MEDIATOR represents the mediator variables, namely, human capital cultivation, innovation resource allocation, and cooperative innovation. Specifically, human capital cultivation is represented as the proportion of employees with master’s and doctoral degrees to the total number of employees (Hu-resource). Innovation resource allocation is measured using two indicators: the ratio of R&D investment to income (R&D) and Total Factor Productivity (TFP). The variable for green cooperative innovation (Cooperate) is represented as a binary variable, taking the value of 1 if the company has engaged in joint green patent applications with other entities and 0 otherwise. The other settings in Equation (3) are consistent with Equation (1), which was presented earlier. Equation (4) is designed similarly to Equation (3), and Equation (5) builds upon Equation (4) by introducing the mediator variables (MEDIATOR). The results of the mediation analysis are shown in

Table 9. Results of mechanism tests.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Hu-Resource	GI	R&D	GI	TFP	GI	Cooperate	GI
DT	0.007 ***	0.052 ***	0.004 ***	0.064 ***	0.074 ***	0.062 ***	0.014 ***	0.055 ***
	(0.001)	(0.011)	(0.001)	(0.011)	(0.009)	(0.011)	(0.004)	(0.009)
Hu-resource		2.945 ***
		(0.353)
R&D				2.079 ***
				(0.315)
TFP						0.149 ***
						(0.021)
Cooperate								1.246 ***
								(0.040)
Constant	0.081 ***	−1.120 ***	0.071 ***	−1.029 ***	4.666 ***	−1.578 ***	−0.285 ***	−0.527 ***
	(0.013)	(0.180)	(0.009)	(0.183)	(0.153)	(0.225)	(0.060)	(0.149)
N	12,837	12,837	12,837	12,837	12,837	12,837	12,837	12,837
CV	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Adj R2	0.251	0.224	0.429	0.202	0.691	0.205	0.092	0.397

Note: robust standard errors clustered at the firm level in parentheses; *** p < 0.01

.

As shown in Table 9, in column (1), the coefficient for DT is 0.07, which is positive and significant at the 1% level. Combined with the estimates in column (2), this suggests that DT promotes GI through human capital cultivation, supporting hypothesis H2. In columns (3) and (5), the coefficients for DT are 0.004 and 0.074, respectively, which are both significant at the 1% level. This shows that DT enhances a company’s ability to allocate innovation resources, validating hypothesis H3 that DT promotes GI through innovation resource allocation. In column (7), the estimated coefficient for DT is 0.014, which is also significant at the 1% level, indicating that DT encourages companies to engage in cross-boundary GI cooperation, thereby enhancing open GI levels. Green cooperative innovation provides companies with diverse knowledge sources and supports GI knowledge and technology. It also improves the overall efficiency of GIC and subsequently, GI performance. Combining these results with those in column (8) indicates that DT promotes GI through enhanced cooperation, thus validating H4.

Based on the estimated results of the mediation model and theoretical analysis, it can be concluded that human capital cultivation, innovation resource allocation, and cooperative innovation are all intermediate mechanisms through which corporate DT impacts GI. This outcome also suggests that in the digital economy era, independent and cooperative innovation are not mutually exclusive. Corporate DT promotes active participation in GI by fostering both internal and external innovation paradigms through open innovation, continually accumulating the relevant knowledge required for GI, and improving GI performance.

7. Discussion

7.1. Findings

With the leapfrog development of human information civilization, the era of Big Data has arrived, and the digital technology represented by Big Data and Cloud Computing is on the rise. DT provides a new idea for enhancing enterprises’ green innovation ability. This study has examined the effect of corporate DT on GI, which is a critical driver of high-quality development in the digital economy era. Leveraging an investigation into the essence and structure of the GIC and utilizing data from Chinese A-share listed companies spanning from 2013 to 2021, this study has yielded several findings. It is important to note that

Table 10. Results of hypothesis testing.

Hypotheses	Contents	Hypothesis Testing Results
H1	Enterprises DT has a positive effect on GI.	Validated
H2	Enterprise DT promotes GI through human capital cultivation.	Validated
H3	Enterprise DT promotes GI by improving the ability to allocate innovative resources.	Validated
H4	Enterprise DT promotes GI by enhancing cooperative innovation capabilities.	Validated

shows that all four of our hypotheses are empirically supported.

(1)

Corporate DT positively affects GI behavior. This conclusion remains robust even after addressing endogeneity concerns and taking robustness tests.

(2)

DT not only augments the quantity of green patent applications but also elevates the quality and efficiency of GI activities within enterprises.

(3)

SEMs, companies receiving government R&D subsidies, facing lower environmental regulatory pressures, and possessing higher-quality information disclosure tend to experience a more pronounced positive impact of DT. Moreover, the effect of DT on GI becomes more potent as industry competition intensifies, although this effect shows diminishing returns beyond a certain threshold.

(4)

DT exerts its influence on GI through three principal mechanisms: the improvement of human capital, the allocation of innovation resources, and the facilitation of cooperative innovation.

Our analysis of the impact of DT on green innovation GI from the perspective of IC aligns with existing research from perspectives such as enterprise value creation, open innovation, technological progress, and biased technological progress. For instance, Gil-Gomez et al. [11], Liu et al. [12], and Shao et al. [22] suggested that DT improves GI performance by reducing management costs, increasing R&D investment, and fostering business model innovation. This is consistent with the macro-technological progress analysis by published literature [17,18,61]. Huan et al. [14] offered a similar perspective by examining the impact of DT on GI in the context of innovation networks within the electric vehicle industry. However, our study extends this analysis by suggesting that the impact of DT on GI is not limited to the automotive sector but is also prevalent across other industries. Furthermore, we refine the concept of innovation networks by focusing on the innovation chain, which emphasizes the vertical structure of innovation from knowledge to products, rather than the broader network structure.

In terms of generalizability, our analysis of Chinese listed companies aligns with studies conducted in other countries. For example, Le et al. [15] found a positive relationship between digitalization and sustainable performance in SMEs in Vietnam’s food and beverage industry. However, there are discrepancies with some existing literature. For instance, Han et al. [23] used machine learning methods to show that the positive effect of DT on GI is mainly observed in non-state-owned, competitive industries, with a non-linear inverted U-shape effect in the overall sample. Yuan et al. [21] found a similar non-linear relationship in agriculture-related enterprises. These inconsistencies may arise from differences in indicator construction, research methods, or industry characteristics, as digitalization levels in agriculture are generally lower than in industrial sectors. These differences warrant further exploration in future research. Nonetheless, our study offers a unique contribution by expanding the mechanism of DT’s impact on GI from the perspective of the IC.

7.2. Implications for Policymakers

Firstly, governments should play a crucial role in facilitating DT by implementing policies that promote both digital industrialization and DT within industries. This support should aim to address practical challenges and obstacles faced by companies during their DT journeys, especially small enterprises. Additionally, governments can accelerate the adoption and application of blockchain technology in GI.

Secondly, companies should not only invest in digital infrastructure and technologies but also enhance their capabilities in managing data elements and other innovation resources. To do so, they can learn from industry leaders and develop tailored DT strategies based on their digital maturity and resource attributes. Companies should also focus on building and optimizing their human capital, as a skilled workforce is critical for GI success.

Thirdly, encouraging collaboration between industry, academia, and research institutions is essential for successful DT in GI. Emphasis should be placed on digitizing key aspects of the GIC, facilitating seamless collaboration between various stakeholders, and providing support for collaborative DT projects. Establishing GI networks and platforms that leverage digital technologies can further drive cooperative efforts.

7.3. Limitations and Future Research

This study has certain limitations that suggest directions for future research: (1) future research can explore green cooperative innovation from multiple dimensions, including input and output measures, as well as factors such as ownership and the establishment of research alliances; (2) while this paper identified three key mechanisms through which DT influences GI, there may be other potential mechanisms to explore; and (3) this study employed a broad sample of A-share listed companies, which may not fully account for industry-specific nuances. Future research could focus on individual industries to gain deeper insights into the impact of DT on GI within specific sectors.