Will the “Underlying Technology” Digital Transformation Promote Substantive Green Innovation in Enterprises?—Evidence from Chinese A-Share Listed Companies

Abstract

Promoting substantive green innovation is the core pathway for enterprises to achieve sustainable development. However, its inherent characteristics of high investment and high risk often result in insufficient innovation motivation among enterprises. Rooted in the Resource-Based View (RBV) and Dynamic Capability Theory (DCT), the research investigates the influence of underlying technology digital transformation on enterprises’ substantive green innovation. Using panel data from Chinese A-share listed firms (2009–2024), this analysis reveals a significant promotional effect of underlying technology digital transformation on substantive green innovation. The robustness of this conclusion is confirmed by a battery of tests. Mechanism analysis demonstrates that this effect functions mainly through two pathways: “technology empowerment” and “governance optimization”, namely enhancing corporate R&D capability and improving ESG performance. Heterogeneity analysis further indicates that this promotional effect is more prominent in enterprises with higher environmental disclosure levels and better internal control quality. This study elucidates the internal mechanism and boundary constraints by which underlying technology digital transformation empowers substantive green innovation, thereby offering micro-level evidence for comprehending the in-depth integration of digital technologies and eco-friendly development. The findings offer important practical implications for firms in formulating effective “digitalization–greenization” synergy strategies.

1. Introduction

The United Nations Environment Programme (UNEP) stated in its 2024 Global Resources Outlook that the global extraction of natural resources has increased threefold over the past five decades. If the current pattern continues, resource extraction by 2060 may increase by another 60% compared to 2020, which will further exacerbate concerns, including climate change, biodiversity loss, and pollution. Against this backdrop, how to achieve the “decoupling” of economic development from resource consumption and promote sustainable economic growth has become a core challenge commonly faced by all countries. Green innovation can not only reduce carbon emission intensity through technological breakthroughs but also foster new growth drivers in emerging industries. It possesses both environmental benefits and economic value and is therefore regarded as a key pathway to promote sustainable development. This trend is highly aligned with the objective of carbon neutrality through clean technologies and circular economy models, as highlighted in international policy frameworks such as the European Green Deal. The Technology Executive Committee (TEC) of the United Nations Framework Convention on Climate Change (UNFCCC) has also clearly pointed out that innovation exerts a vital role in bridging the implementation gap of low-carbon transition. As the largest developing country, China is committed to fulfilling its international responsibilities and has articulated the grand strategic targets of “achieving carbon peaking before 2030 and carbon neutrality before 2060”. However, despite the great significance of green innovation, it is frequently restricted by features including long R&D cycles, high costs, and uncertain short-term returns. As a result, many enterprises lack the internal motivation and capability to carry out substantive green innovation characterized by breakthroughs in core technologies [1].

Meanwhile, digital technologies represented by big data, AI, blockchain, cloud computing, and other technologies are accelerating their iteration. This not only fuels the flourishing development of the digital economy but also makes digital transformation a core driving force for the quality-driven growth of enterprises. It provides a new possibility to overcome the quintessential green innovation dilemma of high investment, long cycle and significant risk. Specifically, relying on the data intelligence technology system, digital transformation enables the precise allocation and dynamic optimization of resource factors. Thereby, it systematically cuts down the energy and carbon footprint per unit of output and boosts overall eco-efficiency. Meanwhile, digital transformation has likewise reshaped the corporate innovation ecosystem. By leveraging the resource integration capabilities of the industrial Internet and the decision-making optimization of intelligent algorithms, it promotes the evolution of innovation activities toward an open and collaborative paradigm, ultimately helping enterprises build sustainable competitive advantages. The COP29 Declaration on Green Digital Actions also explicitly calls for leveraging the enabling role of digital technologies in climate action and building a global framework for the synergy between digitalization and sustainable goals. It can be seen from this that in-depth exploration of the operational mechanisms and enabling paths of digital transformation on green innovation has significant theoretical value and practical implications for energizing green innovation vitality and facilitating the synergistic transformation of “digitalization–greenization”.

The core question addressed in this paper is: How does underlying technology digital transformation serving as the digital foundation impact enterprises’ substantive green innovation? Furthermore, what constitutes the inherent operating mechanisms and contextual boundary conditions of this effect? To address this question, grounded in RBV and DCT, this study employs panel data (Chinese A-shares, 2009–2024) and a fixed-effects model to investigate the direct influence of underlying technology digital transformation on substantive green innovation, and focuses on revealing its inherent “technology–governance” dual-drive mechanism. Specifically, it enhances R&D capabilities through the “technology empowerment” path and improves ESG performance through the “governance optimization” path. The possible marginal contributions of the study are presented below: First, unlike studies on generalized digital transformation, this paper focuses on the in-depth transformation of “underlying technologies” at the level of technological infrastructure and reveals their unique enabling role in substantive green innovation; second, it constructs and empirically tests a dual-path mediating model of “technology empowerment–governance optimization”, systematically clarifies the internal mechanism of digitalization facilitating greenization, and advances the theoretical scholarship on the “digitalization–greenization” synergy transformation; third, it identifies the heterogeneous moderating effects of environmental information disclosure and internal control level, providing a theoretical foundation for enterprises to construct differentiated “digitalization–greenization” synergy strategies.

2. Literature Review and Research Hypothesis

2.1. Substantive Green Innovation

Green innovation, as an important route for firms to achieve sustainable development, fundamentally entails synergistically optimizing environmental and economic benefits through technological innovation, management innovation, and business model innovation [2]. This is specifically reflected in the steady decline in resource consumption intensity, the effective control of pollutant emissions, and the holistic construction of eco-friendly value chains. Compared to traditional innovation, green innovation offers a triple advantage: environmentally, it promotes the “Dual Carbon” Goals through the adoption of clean technology; economically, it cultivates differentiated competitive advantages by enhancing energy efficiency and developing green products; and socially, it establishes stakeholder trust by fulfilling corporate social responsibilities.

At the theoretical level, scholars typically classify green innovation based on two dimensions: motivation and content. From the motivation dimension, it can be categorized into substantive innovation aimed at building long-term competitiveness and strategic innovation focused on meeting policy compliance requirements [3,4]. From the content dimension, it can be categorized into green product innovation that emphasizes product eco-design and green process innovation centered on optimizing production processes [5]. Regarding driving mechanisms, the interplay between external institutional environments, market contexts, and internal organizational factors establishes the key influential framework. At the external level, environmental regulations and fiscal incentives within the “Porter Hypothesis” framework create policy push factors [6,7], while consumer green preferences and media oversight constitute market pull factors [8,9]. Internally, a company’s resource endowments, organizational capabilities, and senior management’s environmental awareness collectively establish the microfoundations for implementing green innovation [10,11,12].

2.2. Underlying Technology Digital Transformation

In this new phase where the digital economy deeply permeates all sectors of society and the economy, digital transformation is driving enterprises to break through traditional growth paths and achieve profound changes in development models [13]. Digital technologies, exemplified by big data and cloud computing, have been deeply integrated into corporate operations. This integration not only effectively enhances production efficiency and breaks down information silos but also significantly strengthens the scientific rigor of management decisions, laying the technological foundation for enterprises to build sustainable competitive advantages. From the theoretical perspective of organizational change, the intrinsic nature of digital transformation resides in the systematic reconstruction of the intrinsic logic underpinning firm value creation. Its core characteristic is manifested as a new driving mechanism based on the “data–algorithm–computing power” technical architecture, propelling enterprises towards synergistic evolution across multiple dimensions, including business models, organizational structures, and factor allocation [14,15].

As practical exploration deepens, academia has gradually identified two distinct paths for digital transformation. The first is underlying technology digital transformation centered on artificial intelligence, blockchain, cloud computing, and big data. This approach focuses on building the underlying technical architecture for data collection, transmission, processing, and intelligent decision-making, serving as the “digital foundation” that drives business system innovation. The second is the digital transformation of “practical application technology”, which focuses on applying specific digital tools and technologies to existing business and management scenarios to achieve process optimization and efficiency gains.

Within the realm of green innovation, ample empirical findings confirm that digital transformation exerts a favorable catalyzing role on green innovation. EI-Kassar and Singh illustrate that big data and predictive analytics technologies boost the implementation of green innovation practices [16]. Research by Alkhatib and Li et al. further confirms that the widespread application of digital technologies can accelerate green development [17,18]. Regarding the underlying mechanisms, digital transformation primarily drives corporate green innovation through two pathways: “information synergy” and “resource empowerment”. Specifically, on one hand, digital transformation accelerates information sharing and integration while alleviating information asymmetry, thereby establishing an efficient collaborative information environment [19,20]. On the other hand, it offers crucial resource support for green innovation by mitigating financial constraints [21], boosting input into innovation resources [22], enhancing knowledge absorption capacity [23], and optimizing resource allocation [24].

However, in real economic scenarios, the facilitative effect of digital transformation on green innovation shows significant heterogeneity. In the regional context, the eastern regions, characterized by well-developed digital infrastructure and an increased level of marketization, demonstrate a significantly stronger propulsive effect of digitalization on greenization than the central and western regions [25]. From an industry perspective, manufacturing—especially heavily polluting sectors—is more likely to achieve green technological breakthroughs through foundational technology empowerment and resource coordination [26]. Meanwhile, traditional service industries face limited effects due to technological adaptability constraints. On the firm level, state-owned and large corporations, which benefit from policy advantages, abundant resources, talent pools, and capital reserves, enjoy more distinct impacts of green innovation driven by digitalization. Conversely, small- and medium-sized firms and private firms are confronted with dual constraints of inadequate infrastructure and governance capabilities [27,28].

Overall, existing literature has demonstrated the beneficial effect of digitalization on greenization and identified heterogeneous characteristics at the levels of region, industry, and enterprise scale, laying a crucial foundation for understanding their relationship. However, several areas require further exploration: first, existing research predominantly conceptualize digital transformation as a holistic concept, overlooking the unique role of underlying technology digital transformation as the foundation and its potential to generate differentiated effects; second, few scholars have explored the promotional effects of digital transformation on substantive green innovation in terms of technological depth, overlooking the unique value of underlying technology digital transformation for substantive green innovation; third, existing research have yet to fully unpack the intrinsic mechanisms and contingent conditions that link digital transformation to substantive green innovation.

With the aim of bridging these research gaps, this research concentrates on the profound transformation of “underlying technologies” at the technological infrastructure level. It empirically examines their distinctive technology empowerment and underlying mechanisms for substantive green innovation and identifies the heterogeneous moderating effects of environmental information disclosure and internal controls. This offers micro-level evidence for firms to align with the low-carbon initiative and formulate “digitalization–greenization” synergy policies.

2.3. Underlying Technology Digital Transformation and Substantive Green Innovation

Against the backdrop of ecological resource constraints and intensifying environmental pollution, substantive green innovation has become a pivotal corporate strategy to break through the bottlenecks of sustainable development. However, its long cycles, high trial-and-error risks, and resource-dependent characteristics have led to a lack of motivation for green innovation and a constraint on its efficiency [1]. Drawing on the RBV [29] and DCT [30], underlying technology digital transformation systematically addresses the resource gaps and efficiency challenges in green innovation by restructuring the logical chain of corporate resource acquisition and integration. At the external resource acquisition level, underlying technology digital transformation leverages digital technologies to considerably extend enterprise information channels and boost information transparency. This not only effectively reduces the costs of searching for external resources but also enhances the accessibility of corporate financing, providing crucial external resource support for green innovation activities [22]. Regarding internal resource integration, underlying technology digital transformation significantly enhances enterprises’ information integration and utilization efficiency, facilitating cross-departmental information flow and resource sharing [31]. This process achieves efficient allocation of production factors, significantly reduces trial-and-error costs, and effectively avoids resource wastage caused by redundant work. The dual optimization mechanism of external resource acquisition and internal resource integration continuously supports enterprises in enhancing their green innovation performance. Given the foregoing analysis, the study advances the following hypothesis:

H1. 

Underlying technology digital transformation positively promotes substantive green innovation.

3. Data and Research Design

3.1. Sample and Data Sources

Considering data availability, this research initially selected the full sample of Chinese A-share listed firms over 2009–2024. In line with standard research practice, the research introduced the following modifications: first, samples of firms classified as ST, *ST, PT, and those in the financial industry during the sample period were eliminated; second, to mitigate information disclosure disparities resulting from short listing periods, samples of companies listed in the same year were removed; third, observations with incomplete or extreme values in the core variables were omitted; finally, to account for outlier effects, continuous variables were subjected to trimmed quantile scaling at the 1st and 99th percentiles. Following these procedures, the study ultimately yielded 37,023 valid observations. In terms of data sources, green patent data originated from the CNRDS; digital transformation text analysis data were derived from annual reports on the Giant Tide Information Network; personnel structure information and ESG ratings were retrieved from the Wind Information; all other relevant data originated from the CSMAR Database.

3.2. Variable Description

Dependent variable: Substantive green innovation (Greinva). In line with Qi et al. [32], this research employs ln (the annual count of a firm’s independently filed green invention patents + 1) as the measurement.

Independent variable: Underlying technology digital transformation (Digital). Drawing upon the digital feature word map constructed by Wu et al. [33], this study centers on “underlying technologies” at the level of technological infrastructure. In terms of four dimensions—artificial intelligence, big data, cloud computing, and blockchain—this study utilized Python 3.11.2 to capture the narrative disclosures of listed firms’ annual reports. Feature words related to digitalization were identified, and their frequencies were counted. The annual total word frequency was calculated by aggregating the frequencies across all dimensions. To alleviate the influence of right-skewed distribution, ln (total frequency + 1) was taken to form the measurement.

Control variables: Drawing from existing scholarly research, a set of potential influencing factors on substantive green innovation are designated as controls, which include: enterprise size (Size), enterprise age (Age), management shareholding percentage (Share), proportion of independent directors (Indep), asset–liability ratio (Lev), long-term debt ratio (Ldr), book-to-market ratio (Bm), audit opinion (Aud), CEO duality (Dual), and nature of ownership (Soe). To control for year and industry variations, dummy variables for both year and industry (Year and Industry) are incorporated. The definitions and measurements of all variables are presented in

Table 1. Variable definitions.

Variable Type	Variable Name	Variable Symbol	Variable Description
Dependent Variable	Substantive Green Innovation	Greinva	Ln (Total Green Patent Applications + 1)
Independent Variable	Underlying Technology Digital Transformation	Digital	Ln (Underlying Technology Digital Transformation Keyword Frequency + 1)
Control Variables	Enterprise Scale	Size	Ln (Total Assets)
	Enterprise Age	Age	Ln (Age of Establishment + 1)
	Management Shareholding Percentage	Share	Number of Shares Held by Executives/Total Shares Issued
	Proportion of Independent Directors	Indep	Number of Independent Directors/Total Number of Directors
	Asset–Liability Ratio	Lev	Total Liabilities/Total Assets
	Long-Term Debt Ratio	Ldr	Long-Term Liabilities/Total Liabilities
	Book-to-Market Ratio	Bm	Book Value of Shareholders’ Equity at Year-End/Market Value
Binary Control Variables	Audit Opinion	Aud	A dummy variable that equals 1 if the audit opinion is an unqualified opinion, and 0 if not
	CEO Duality	Dual	A dummy variable for CEO duality, which equals 1 if the roles of CEO and board chair are held by the same individual, and 0 if not
	Nature of Ownership	Soe	A dummy variable for state ownership, which equals 1 if the firm is a state-owned enterprise, and 0 if not
Fixed Effects	Year	Year	Time dummy variable
	Industry	Industry	Industry dummy variable

Note: total assets are denominated in CNY.

.

3.3. Model Specification

For the purpose of investigating the implications of underlying technology digital transformation for substantive green innovation, model (1) was developed:

$$G_{i,t}={\alpha }_0+{\alpha }_1{D}_{i,t}+\sum {\alpha }_j{CVs}_{j,i,t}+\sum \lambda +\sum Year+\sum Ind+{\epsilon }_{i,t}$$

(1)

Here, i, t denotes the sample data for the i-th individual in year t. G represents substantive green innovation (Greinva), D signifies underlying technology digital transformation (Digital), CVs are the control variables; λ, Year, and Ind capture the individual, time, and industry fixed effects, respectively; and ε is the random disturbance term.

4. Regression Results Analysis

4.1. Descriptive Statistics and Correlation Analysis

As a concise summary of the main characteristics of the dataset, the descriptive statistics for all relevant variables are presented in

Table 2. Descriptive statistics.

Variable Name	Sample Size	Mean	Median	Standard Deviation	Minimum	Maximum
Greinva	37,032	0.245	0.000	0.617	0.000	3.219
Digital	37,032	1.007	0.693	1.282	0.000	4.836
Size	37,032	22.341	22.149	1.284	19.990	26.320
Age	37,032	2.966	2.996	0.335	1.946	3.638
Share	37,032	0.125	0.005	0.185	0.000	0.668
Indep	37,032	0.376	0.364	0.053	0.333	0.571
Aud	37,032	0.970	1.000	0.169	0.000	1.000
Lev	37,032	0.446	0.443	0.202	0.059	0.911
Ldr	37,032	0.157	0.096	0.173	0.000	0.708
Bm	37,032	0.331	0.309	0.159	0.035	0.792
Dual	37,032	0.283	0.000	0.451	0.000	1.000
Soe	37,032	0.364	0.000	0.481	0.000	1.000

. The study sample comprises 37,032 observations. Core variables (Greinva, Digital) generally exhibit right-skewed distributions, indicating that most sample values are low, and a few high-value samples elevate the average. Regarding corporate characteristics: enterprise size (mean Size = 22.341, i.e., mean Total Assets = 5.041 × 10⁹) aligns with the typical distributions among listed companies, management shareholding percentage (Share) differs widely spanning 66.8%, audit opinions (Aud) are predominantly standard unqualified opinions accounting for 97% of the total. Approximately 28.3% of the sample firms exhibit CEO duality (Dual), while about 36.4% are state-owned enterprises (Soe). Correlation analysis revealed a significant positive correlation between Digital and Greinva, and it passed the correlation tests. The results initially reveal that underlying technological digital transformation exerts a positive promotional impact on substantive green innovation. Multicollinearity tests conducted on all variables showed that the variance inflation factors (VIFs) were consistently below 10, indicating no multicollinearity among the variables. This verifies that the model satisfies the fundamental assumptions required for regression analysis.

4.2. Baseline Regression

Table 3. Baseline regression results.

	(1) Greinva	(2) Greinva	(3) Greinva
Digital	0.0483 ***	0.0197 ***	0.0221 ***
	(8.94)	(3.89)	(4.30)
_cons	0.197 ***	−1.050 ***	−0.919 ***
	(36.17)	(−6.08)	(−2.97)
Control	No	Yes	Yes
Year/Industry	No	No	Yes
N	37,032	37,032	37,032
F	79.930 ***	11.890 ***	4.001 ***
Adj-R2	0.622	0.626	0.628

Note: *** p < 0.01.

displays the baseline regression results pertaining to the influence of underlying technology digital transformation (Digital) on substantive green innovation (Greinva). The term “_cons” represents the constant term (intercept) of the regression model. Column (1) shows that when controlling only for individual fixed effects, the coefficient for Digital stands at 0.0483 and is significant at the 1% level. Column (2), which incorporates a series of control variables on the basis of Column (1), indicates the coefficient for Digital declining to 0.0197 while remaining highly significant (p < 0.01). This indicates that omitted variables may overestimate the contribution of underlying technology digital transformation. Column (3) demonstrates that, with the addition of year and industry fixed effects, the conclusion remains virtually unchanged. An evaluation of economic significance indicates that a one-unit increase in underlying technology digital transformation results in, on average, a 2.21% rise in substantive green innovation. The adjusted R² reaches 0.628, and the F statistic is significant, confirming the model’s validity and strong explanatory power. Thus, research hypothesis H is validated.

4.3. Robustness Analysis

For the purpose of result validation, multiple robustness tests were implemented, ranging from the instrumental variable and Heckman two-stage methods to altering variables, methods, and sample scope.

4.3.1. Instrumental Variables Approach

To tackle potential endogeneity biases—including the reverse causality in which enterprises with more robust green innovation capabilities tend to proactively promote underlying technology digital transformation, as well as the omitted variable issue where unobservable factors simultaneously influence both corporate digital decision-making and green innovation practices—this study utilizes the instrumental variables approach for re-examination. Following the instrumental variable construction method proposed by Zhang and Zhao [34], this study constructs the instrumental variable (IV) as “National E-commerce Demonstration Cities dummy variable × mean Digital of enterprises in the same city, same industry, and same year”. The “National E-commerce Demonstration City” policy represents an exogenous shock driven by the central government, with pilot city selection based on macro regional economic indicators rather than firm-specific characteristics, thereby satisfying the exogeneity requirement for an instrumental variable. At the same time, the core objective of this policy is to accelerate the advancement of e-commerce and the digital economy, which effectively drives corporate digital transformation and meets the relevance requirement between the instrumental variable and the endogenous variable [35].

The results of the endogeneity test are displayed in

Table 4. Instrumental variables estimates.

	(1) Digital	(2) Greinva
IV	0.553 ***
	(0.019)
Digital		0.031 *
		(0.017)
Control	Yes	Yes
Year/Industry	Yes	Yes
Kleibergen–Paap rk LM Statistic	410.001 [0.000]
Kleibergen–Paap rk Wald F Statistic	840.921 [16.380]
N	37,032	37,032
Adj-R2	0.7938	0.6272

Note: * p < 0.1, *** p < 0.01.

. First, IV is positively correlated with Digital (first-stage coefficient = 0.553, p < 0.01), fulfilling the correlation requirement. Second, the Kleibergen–Paap rk LM statistic (LM = 410.001, p < 0.01) rejects the null hypothesis of “non-identifiability”. Furthermore, the Kleibergen–Paap Wald rk F statistic (F = 840.921) far exceeds the 10% critical value of 16.380 for the Stock–Yogo weak identification test, eliminating the problem of weak instrumentation. Taken together, the instrumental variable passes the relevant diagnostic tests, confirming its validity. Regression findings indicate that after addressing endogeneity, the promotional effect of underlying technology digital transformation on substantive green innovation holds at the 10% significance level, demonstrating robust research conclusions.

4.3.2. Heckman Two-Stage

A potential sample self-selection bias warrants consideration. Specifically, enterprises engaged in digital transformation may exhibit inherent systemic characteristics (e.g., stronger resource endowments or innovation-oriented cultures), and their green innovation performance could be inherently superior to that of non-transforming enterprises, thereby introducing confounding effects into the research conclusions. To alleviate this endogeneity issue, this study incorporates the Heckman two-stage approach within its analytical framework for additional validation. The detailed procedure is outlined below: in the first stage, a dummy variable Dig_dum is constructed to indicate whether a firm has implemented underlying technology digital transformation (it equals 1 if yes, and 0 if not); then, the average level of underlying technology digital transformation among other firms in the same industry and year (Dig_mean) is calculated; next, a Probit regression is conducted with Dig_dum as the dependent variable, and Dig_mean along with the original control variables as explanatory variables; finally, the Inverse Mills Ratio (IMR) is constructed using the estimates obtained from the Probit model; in the second stage, IMR is incorporated into the baseline model as an additional control variable and the regression is run with it alongside Digital and other controls. Given the difficulty of incorporating individual fixed effects into standard Probit models, this study merely accounted for year and industry fixed effects in the first stage. For the second stage, individual, year, and industry fixed effects were simultaneously controlled.

As demonstrated in Column (1) of

Table 5. Heckman Two-Step estimates.

	The First Stage	The Second Stage
	(1) Dig_dum	(2) Greinva
Dig_mean	0.119 **
	(2.06)
Digital		0.0167 ***
		(2.68)
IMR		0.267 ***
		(3.39)
_cons	−5.514 ***	−1.678 ***
	(−26.83)	(−3.75)
Stkcd	No	Yes
Control	Yes	Yes
Year/Industry	Yes	Yes
N	37,032	37,032
Pseudo R2	0.2560
Adj-R2		0.662

Note: ** p < 0.05, *** p < 0.01. “Stkcd” denotes stock code, representing individual (firm) fixed effects in the model.

, Dig_mean exhibits a significant positive relationship on Dig_dum (p < 0.05), which justifies its selection as an exogenous variable. The Pseudo R² of 0.2560 indicates a reasonably good overall fit. Column (2) shows that the coefficient of Digital stands at 0.0167, significantly positive at the 1% level. This validates that the promotional influence of underlying technology digital transformation on substantive green innovation holds true after correcting for sample self-selection.

4.3.3. Replace Regression Methods

Given that the dependent variable substantive green innovation exhibits a distribution with a high frequency of zero values, and the Ordinary Least Squares (OLS) estimation and its logarithmic transformation used in the benchmark regression may not adequately address the potential estimation bias arising from this characteristic, this study further employs the Tobit, Poisson Pseudo-Maximum Likelihood (PPML), and zero-inflated negative binomial (ZINB) model for robustness checks. Specifically, the Tobit model is applicable for data with a left-censoring point at zero. The PPML and ZINB models, on the other hand, can directly estimate non-negative integer-type patent count data without requiring logarithmic transformation, thereby better handling the prevalence of zero values.

Estimates from the Tobit model are presented in

Table 6. Results of alternative regression methods.

	(1) Greinva	(2) NGreinva	(3) NGreinva
Digital	0.382 ***	0.291 ***	0.385 ***
	(14.77)	(4.47)	(10.57)
_cons	−13.32 ***	−21.50 ***	−16.42 ***
	(−15.28)	(−8.99)	(−13.17)
Stkcd	No	No	No
Control	Yes	Yes	Yes
Year/Industry	Yes	Yes	Yes
N	37,032	36,919	37,032
F	57.468 ***
Pseudo R2	0.0745	0.3840

Note: *** p < 0.01.

, Column (1). After accounting for the censored nature of the dependent variable, the coefficient for Digital remains significantly positive (α₁ = 0.382, p < 0.01). Column (2) reports the results from the PPML model, where the dependent variable is the original patent count (NGreinva). The estimated coefficient remains significant (α₁ = 0.291, p < 0.01). Column (3) displays the ZINB results, with Digital showing a significant positive coefficient (α₁ = 0.385, p < 0.01). These findings are consistent with the baseline OLS estimates and further corroborate research hypothesis H.

4.3.4. Other Robustness Analysis

(1) Replace the dependent variable. Given the notable discrepancy between applications and grants, the research substituted the dependent variable with the natural logarithm of green patent grants plus one (denoted as Greinva_1) for robustness analysis.

Table 7. Other robustness analysis results.

	Replace the Dependent Variable	Replace the Independent Variable		Change Sample Capacity
	(1) Greinva_1	(2) Greinva	(3) Greinva	(4) Greinva
Digital	0.0188 ***			0.0214 ***
	(4.68)			(3.06)
Digital_1		0.0202 ***
		(3.87)
Digital_2			0.440 ***
			(3.55)
_cons	−0.829 ***	−0.938 ***	−0.964 ***	−0.900 **
	(−3.88)	(−3.02)	(−3.15)	(−2.56)
Stkcd	Yes	Yes	Yes	Yes
Control	Yes	Yes	Yes	Yes
Year/Industry	Yes	Yes	Yes	Yes
N	37,032	37,032	36,846	19,948
F	6.291 ***	3.699 ***	3.305 ***	2.331 ***
Adj-R2	0.551	0.627	0.626	0.646

Note: ** p < 0.05, *** p < 0.01.

shows that after replacement, the coefficient for Digital remains significantly positive (α₁ = 0.0188, p < 0.01), again validating research hypothesis H.

(2) Replace the independent variable. This study substituted the core independent variable with two alternative measures. The first alternative is calculated as the natural logarithm of the total word frequency of “underlying technology” characteristics in Management’s Discussion and Analysis (MD&A) texts plus one (denoted as Digital_1), and the second is computed as the percentage of digital-related word frequency of “underlying technology” in MD&A texts multiplied by 100 (i.e., 100 × word frequency of “underlying technology” digital characteristics/total length of MD&A text, denoted as Digital_2). Then, it conducted regression again. Table 7 indicates that the positive association for Digital continues to show statistical significance after replacement, further validating the robustness of research hypothesis H.

(3) Change sample capacity. To eliminate potential interference from the major exogenous shock of the 2020 COVID-19 pandemic on the estimates, this research excluded all samples from 2020 onwards and reran the regression. Column (4) of Table 7 shows that the coefficient for Digital remains significantly positive (α₁ = 0.0214, p < 0.01), demonstrating that the core findings are robust and unaffected by samples from exceptional periods.

In summary, following a battery of robustness tests, the regression results still offer robust support for the proposed hypotheses. Therefore, the core findings have been validated by the robustness tests, demonstrating good robustness.

5. Further Analysis

5.1. Mechanism Analysis

5.1.1. “Technology Empowerment” Pathway: Strengthening R&D Capability

Based on DCT [30], the underlying technology digital transformation provides in-depth technological empowerment to corporate R&D activities by restructuring organizational practices and resource bases. For one thing, digital technologies drive an increase in R&D investment to respond to digital environmental changes. For another, underlying technologies such as big data significantly streamline knowledge management procedures, elevating enterprises’ efficiency in capturing, integrating, and transferring complex cross-boundary information during green technology exploration [36]. R&D capability serves as the micro-foundation for enterprises to overcome the high-cost, high-risk barriers to substantive green innovation. Firms with strong R&D capabilities proactively seize strategic opportunities to surmount innovation barriers, while those with weak capabilities often suppress green innovation activities due to resource constraints and risk-averse tendencies [37,38]. Concurrently, enhanced R&D capability directly empowers the optimization of the production process, effectively lowering energy use and emissions to achieve energy-saving and carbon-reduction targets [39]. Thus, underlying technology digital transformation can enhance corporate R&D capability through technology empowerment, thereby promoting substantive green innovation.

To examine the mediating effect of R&D capability (Rd), while also enhancing the temporal logic of the mechanism analysis and mitigating potential endogeneity bias, the following lagged mediation model was constructed for testing:

$$\begin{array}{c}Rd={\beta }_0+{\beta }_{1}L.D+\sum {\beta }_j{CVs}_{j,i,t}+\sum \lambda +\sum Year+\sum Ind+{\epsilon }_{i,t}\end{array}$$

(2)

$$\begin{array}{c}{G}_{i,t}={\gamma }_0+{\gamma }_1{L.D}_{i,t}+{\gamma }_{2}Rd+\sum {\gamma }_j{CVs}_{j,i,t}+\sum \lambda +\sum Year+\sum Ind+{\epsilon }_{i,t}\end{array}$$

(3)

In the above model, the independent variable underlying technology digital transformation (Digital) is lagged by one period (t − 1) to predict the current-period mediator variable, R&D capability (Rd), and then to examine its impact on the current-period substantive green innovation (Greinva). Here, L.D denotes the one-period lagged underlying technology digital transformation (L.Digital), Rd represents the enterprise’s R&D capability. Following the research by Liu et al. [40], this study measures R&D capability using the sum of the Min–Max standardized values for five ratios: R&D expenditure/total assets, R&D investment/operating revenue, R&D personnel/total employees, technical personnel/total employees, and employees with master’s degrees or higher/total employees. A higher value indicates stronger R&D capability. Column (1) of

Table 8. Mechanism Analysis Results.

	(1) Rd	(2) Greinva	(3) ESG	(4) Greinva
L.Digital	0.00883 ***	0.0231 ***	0.0442 ***	0.0240 ***
	(6.58)	(3.90)	(3.79)	(4.01)
Rd		0.154 ***
		(3.49)
ESG				0.0104 ***
				(2.92)
_cons	−0.147	−0.816 **	−5.677 ***	−0.780 **
	(−1.63)	(−2.19)	(−8.22)	(−2.11)
Control	Yes	Yes	Yes	Yes
Year/Industry	Yes	Yes	Yes	Yes
N	29,365	29,365	29,365	29,365
F	9.341 ***	3.048 ***	52.120 ***	3.013 ***
Adj-R2	0.862	0.640	0.428	0.640

Note: ** p < 0.05, *** p < 0.01.

reports the regression coefficient between L.Digital and Rd is 0.00883 (p < 0.01). This finding is consistent with the theoretical expectation that underlying technology digital transformation contributes to enhancing firms’ subsequent R&D capability. Column (2) indicates that, after controlling for R&D capability, L.Digital still exhibits a significant positive association with Greinva (γ₁ = 0.0231, p < 0.01), while the coefficient for Rd also remains significantly positive (γ₂ = 0.154, p < 0.01). These statistical characteristics are consistent with the criteria for partial mediation, suggesting that underlying technology digital transformation may promote substantive green innovation by empowering firms’ R&D capability, thereby providing support for the corresponding mechanism.

5.1.2. “Governance Optimization” Pathway: Enhancing ESG Performance

Underlying technology digital transformation enhances ESG (environmental, social, and governance) performance through governance optimization pathways, thereby cultivating a supportive organizational ecosystem for substantive green innovation. Digital transformation achieves governance optimization through three mechanisms: first, digital technologies mitigate information asymmetry, strengthen stakeholder oversight, and increase the cost of corporate information falsification [41,42], thereby curbing strategic information disclosure and promoting governance transparency; second, digital technologies optimize internal information flow and risk management mechanisms, enhancing decision-making rigor and thereby elevating corporate social responsibility fulfillment; third, digital technologies enable precise monitoring and management of resource consumption and pollution emissions, directly improving corporate environmental responsibility performance. Outstanding ESG performance not only stems from high-quality governance but also provides critical resources for long-term, high-uncertainty substantive green innovation by alleviating financing constraints, enhancing corporate reputation, and securing policy support [43]. Thus, underlying technology digital transformation can facilitate corporate ESG performance through governance optimization, thereby promoting substantive green innovation.

To investigate the mediating effect of ESG performance (ESG), while also enhancing the temporal logic of the mechanism analysis and mitigating potential endogeneity bias, the following lagged mediation model was constructed for testing:

$$\begin{array}{c}ESG={\beta }_0+{\beta }_{1}D+\sum {\beta }_j{CVs}_{j,i,t}+\sum \lambda +\sum Year+\sum Ind+{\epsilon }_{i,t}\end{array}$$

(4)

$$\begin{array}{c}{G}_{i,t}={\gamma }_0+{\gamma }_1{D}_{i,t}+{\gamma }_{2}ESG+\sum {\gamma }_j{CVs}_{j,i,t}+\sum \lambda +\sum Year+\sum Ind+{\epsilon }_{i,t}\end{array}$$

(5)

In the above model, the independent variable underlying technology digital transformation (Digital) is lagged by one period (t − 1) to predict the current-period mediator variable, ESG performance (ESG), and then to examine its impact on the current-period substantive green innovation (Greinva). Here, L.D denotes the one-period lagged underlying technology digital transformation (L.Digital), ESG represents corporate ESG performance. This paper employs the Gao et al. Approach [44], where the Huazheng ESG Rating Index (with 9 tiers from C to AAA) is converted to a score on a 1 to 9 scale. Column (3) of Table 8 shows that the regression coefficient between L.Digital and ESG is 0.0442 (p < 0.01). This finding conforms to the theoretical expectation that underlying technology digital transformation helps improve a firm’s subsequent ESG performance. Column (4) indicates that, after controlling for ESG performance, L.Digital still exhibits a significant positive association with Greinva (γ₁ = 0.0240, p < 0.01), while the coefficient for ESG also remains significantly positive (γ₂ = 0.104, p < 0.01). These statistical characteristics are consistent with the criteria for partial mediation, suggesting that underlying technology digital transformation may promote substantive green innovation by enhancing a firm’s ESG performance, thereby providing support for the corresponding mechanism.

5.2. Heterogeneity Analysis

5.2.1. Environmental Information Disclosure

Environmental information disclosure essentially serves as a strategic resource signal transmitted by enterprises to external markets. By providing high-quality environmental disclosure, firms can mitigate information gap with external stakeholders, which in turn effectively relieves financing constraints and raises their resource acquisition capacity. Simultaneously, enterprises with higher disclosure levels are better aligned with government expectations, thereby gaining access to policy support measures like tax incentives. Drawing on the preceding analysis, this study argues that underlying technology digital transformation exerts a particularly potent promotional effect on substantive green innovation within firms characterized by more robust environmental disclosure. In line with the research of Zhang and Dong [45], the study measures environmental information disclosure (EID) by aggregating the scores of all secondary indicators from the five-dimensional framework in the CSMAR database. Enterprises with annual EID values exceeding the median are categorized as the high-disclosure group (EID = 1), while the rest are assigned to the low-disclosure group (EID = 0). This study employs both subgroup regression and interaction term regression for testing, and constructs the following model for interaction term regression analysis:

$$\begin{array}{c}{G}_{i,t}={\lambda }_0+{\lambda }_1{D}_{i,t}+{\lambda }_{2}EID+{\lambda }_{3}D\times EID+\sum {\lambda }_j{CVs}_{j,i,t}+\sum \lambda +\sum Year+\sum Ind+{\epsilon }_{i,t}\end{array}$$

(6)

Here, EID denotes the level of environmental information disclosure, while D × EID refers to the interaction term between Digital and EID. Other symbols are defined as above. Higher environmental disclosure levels amplify the favorable contribution of underlying technology digital transformation to substantive green innovation. Thus, the coefficient λ₃ preceding the interaction term is expected to be positive. The detailed regression outcomes are summarized in

Table 9. Results of heterogeneity analysis.

	Environmental Information Disclosure			Internal Control Level
	Subgroup Regression		Interaction Term Regression	Subgroup Regression		Interaction Term Regression
	(1) Low	(2) High	(3)	(4) Low	(5) High	(6)
Digital	0.00849	0.0376 ***	0.0220 ***	0.00599	0.0361 ***	0.0224 ***
	(1.61)	(4.25)	(4.34)	(1.12)	(4.42)	(4.35)
EID			−0.0130 *
			(−1.68)
Digital × EID			0.0317 ***
			(4.27)
DIB						0.00487
						(0.98)
Digital × DIB						0.00819 *
						(1.83)
_cons	−0.873 ***	−0.850	−0.906 ***	−0.534	−1.105 **	−0.881 ***
	(−3.04)	(−1.39)	(−2.97)	(−1.42)	(−2.44)	(−2.86)
Control	Yes	Yes	Yes	Yes	Yes	Yes
Year/Industry	Yes	Yes	Yes	Yes	Yes	Yes
N	18,446	17,438	37,032	17,937	17,775	37,032
F	3.217 ***	3.708 ***	4.175 ***	1.959 **	3.194 ***	3.441 ***
Adj-R2	0.603	0.660	0.628	0.616	0.647	0.628
Fisher’s permutation test p-value	0.000			0.000

Note: * p < 0.1, ** p < 0.05, *** p < 0.01.

. As shown in Columns (1) and (2) for the subgroup analyses, Digital exhibits significantly positive coefficients in both cases. However, the coefficient in the high-disclosure group is larger and shows stronger statistical significance. Moreover, the empirical p-value is 0.000, confirming a more pronounced effect of Digital on Greinva among firms with higher levels of environmental information disclosure. Furthermore, Column (3) reveals significantly positive coefficients for both Digital and its interaction term, further confirming that the favorable influence of Digital on Greinva becomes more substantial as firms’ environmental information disclosure increases.

5.2.2. Internal Control Level

A company’s internal control level is a critical factor in its resource allocation and coordination capabilities. High-quality internal control systems can effectively monitor and optimize the allocation and utilization of resources in digital projects, ensuring that capital and human resources are accurately directed towards green technology R&D activities with long-term environmental benefits. Simultaneously, a robust internal control system enhances a company’s risk identification and assessment capabilities, enabling it to proactively monitor various risks during the green technology R&D process and significantly reduce the probability of green innovation failure caused by uncertainty. The foregoing analysis suggests that the enabling influence of underlying technology digital transformation on substantive green innovation is more pronounced in firms where internal controls are of higher quality. The measurement of internal control relies on the DIB Internal Control Index. Subsequent division of the sample at the median into high-control groups (DIB = 1) and low-control groups (DIB = 0) allows for comparative regression analysis. An interaction term regression model is constructed as follows:

$$\begin{array}{c}{G}_{i,t}={\lambda }_0+{\lambda }_1{D}_{i,t}+{\lambda }_{2}DIB+{\lambda }_{3}D\times DIB+\sum {\lambda }_j{CVs}_{j,i,t}+\sum \lambda +\sum Year+\sum Ind+{\epsilon }_{i,t}\end{array}$$

(7)

Here, DIB denotes the level of corporate internal control, while D × DIB denotes the interaction term between Digital and DIB. Other symbols are defined as above. A higher level of internal control enhances the favorable contribution of underlying technology digital transformation on substantive green innovation. Therefore, the coefficient λ₃ preceding the interaction term is expected to be positive. The detailed regression outcomes are summarized in Table 9. The subgroup analyses in Columns (4) and (5) demonstrate that the coefficients of Digital remain significantly positive in both cases. The high internal control subgroup shows a larger coefficient magnitude and higher statistical significance. Furthermore, the empirical p-value is 0.000, indicating that high-quality internal control systems strengthen the positive effect between Digital and Greinva. Column (6) shows that both Digital and its interaction term have significantly positive coefficients, further indicating that stronger internal controls amplify the influence of Digital on Greinva.

6. Discussion and Conclusions

Based on RBV and DCT, this research leverages a fixed-effects model to assess both the direct influence and the transmission pathways through which underlying technology digital transformation promotes substantive green innovation. Utilizing panel data from Chinese A-share listed firms (2009–2024), the following conclusions are drawn. First, underlying technology digital transformation significantly promotes substantive green innovation. The conclusion remains robust after a range of robustness analysis, including the instrumental variable method, the Heckman two-stage method, replacement of research methods and variables, and adjustment of sample capacity. Second, underlying technology digital transformation promotes substantive green innovation through two pathways: “technology empowerment” and “governance optimization”. Specifically, it enhances firms’ hard capabilities for green technological breakthroughs by strengthening R&D capability, while simultaneously improving their internal and external governance environments for green innovation by enhancing ESG performance. Third, firms with superior environmental information disclosure and stronger internal controls are able to leverage digital transformation more effectively to drive substantive green innovation.

This study demonstrates that underlying technology digital transformation possesses strategic value for facilitating corporate substantive green innovation through the “technology–governance” duality pathway. These results provide insights for motivating enterprises to actively achieve the “Dual Carbon” Goals and formulate digitalization–greenization synergy strategies. First, enterprises should recognize that digital transformation constitutes not only technological upgrading but also a profound governance model shift. Strategic coordination of technological empowerment and governance optimization is required to systematically integrate digital technologies throughout the entire green innovation process. Second, at the technology empowerment level, enterprises should utilize foundational technologies to boost the intensity of green innovation R&D investment, allocating resources preferentially to green technologies that yield substantial environmental benefits. Third, at the governance optimization level, enterprises should establish a dynamic collaborative governance framework that is aligned with digital transformation and green innovation strategies, enhance internal control systems, and refine environmental information disclosure mechanisms. This involves proactively disclosing management measures to prevent “greenwashing” and third-party verification information to strengthen data credibility.