Dual Mechanisms of Digital Transformation in Sustaining Green Innovation: A Supply Chain Perspective on Capability–Motivation Dynamics

Abstract

In the context of global industrial chain decarbonization, the perpetuation of corporate green innovation has emerged as a linchpin for sustaining a competitive advantage in the pursuit of environmental sustainability. Employing a panel data framework, this investigation analyzes A-share listed firms in China from 2011 to 2023. In terms of supply chain perspectives, this study utilizes fixed effects models, mediation analysis, and moderation analysis to empirically examine how downstream enterprises’ digital transformation affects the sustainability of upstream enterprises’ green innovation, while deconstructing the “capability–motivation” dual pathway underlying such sustainability. The key findings are as follows: (1) downstream digital transformation significantly strengthens upstream green innovation persistence through both capability reinforcement and motivation amplification, with a notably stronger impact on the latter; (2) mechanism tests show that capability improvement primarily arises from knowledge spillovers and enhanced supply–demand coordination efficiency, while motivation enhancement stems from intensified market competition and greater responsiveness to tax incentives; and (3) supply chain structural characteristics exert critical moderating effects. This research elucidates the operational logic and boundary conditions of supply chain digital coordination in driving green innovation persistence, contributing to theoretical frameworks while offering actionable insights for policymaking and corporate strategic optimization in sustainable supply chain management.

1. Introduction

In the context of the global response to climate change challenges and the promotion of sustainable development, fostering sustainable supply chain development has emerged as a critical way for companies to strengthen their core competitive edge and realize social value. This perspective has gained widespread recognition within the academic community [1,2,3]. Traditional supply chains are often caught in the predicament of “local environmental compliance and overall unsustainability” due to fragmented information and inefficient synergy [4]. Due to the divergence in objective functions across supply chain segments [5], upstream suppliers’ short-term emission reduction efforts often fail to align with downstream customers’ long-term green demands. This mismatch frequently leads to resource wastage and excessive carbon emissions throughout the entire chain, underscoring the urgent need for supply chain sustainability to be “long-term, collaborative, and systemic”. The fast advancement of digital technology has offered a novel solution to address this predicament: artificial intelligence, blockchains, the Internet of Things, and other tools are reshaping the value creation mode of supply chains by breaking information barriers, optimizing resource allocation, and strengthening process control. It is pushing the sustainability goal from “fragmented practice” to “whole-chain synergy”. The proliferation of emerging technologies like the Internet of Things, blockchains, and quantum computing has intensified the complexity of supply chain conflicts while simultaneously providing stronger support for green supply chain management [6,7].

Despite the potential environmental benefits of digital transformation, how enterprises can effectively transform digital capabilities into sustained green innovation is still a key topic that needs to be investigated. Current research is mostly centered on the digital transformation of core enterprises themselves yet pays little attention to its ripple effects on upstream and downstream supply chain partners. Meanwhile, existing studies largely stop at analyzing whether green innovation occurs or not, without offering a systematic explanation for how to overcome short-termism and establish a long-term, stable iterative mechanism. From the perspective of dynamic capability theory, this is essentially the process where enterprises utilize digital technologies to reconstruct their resource integration and innovation iteration capabilities [8]. Institutional theory suggests that the long-term viability of green innovation is contingent upon the long-term transmission of coercive and imitative pressures within supply chains, along with normative isomorphism [9]. At the end of 2018, Xiaomi Group established a strategic partnership with upstream suppliers TCL Science and Technology Group Company Limited through the Xiaomi ecosystem’s Internet of Things (IoT), accelerating the iteration and upgrading of TCL’s products with the help of digital tools. TCL’s product enhancement through digital technologies has not only given rise to innovative smart products but has also facilitated broader green innovation resource acquisition across supply chain networks. This case demonstrates how digital transformation can integrate supply chain resources and sustain green innovation, offering a practical framework for analyzing the “digital transformation–green innovation” linkage. Green innovation serves as the cornerstone of supply chain sustainability, representing not isolated corporate actions but a systemic process of continuous technological advancement and operational optimization across the entire supply chain. By fostering resource efficiency, environmental responsibility, and social impact, its sustainability determines whether supply chains can transcend short-term compliance and achieve holistic, long-term sustainable operations.

This study investigates how downstream enterprises’ digital transformation influences upstream suppliers’ green innovation sustainability. Utilizing A-share listed company data (2011–2023), we analyze customer-driven digital transformation’s role in sustaining suppliers’ green innovation motivation. By mapping inter-firm supply chain relationships, we further examine the underlying mechanisms and interaction effects between these dynamics. This study is specifically designed to investigate the following research questions: Does downstream digital transformation enhance the green innovation sustainability of upstream enterprises and how? Are there any differences in the specific paths of this enhancement effect on the dimensions of sustainability and sustainability dynamics? How do supply chain concentration characteristics modulate the dual mechanism of downstream digitalization spillovers on upstream green innovation sustainability? By addressing the aforementioned issues, this paper anticipates making threefold contributions at both theoretical and practical levels. Theoretically, it establishes a conceptual framework that connects digital transformation to the sustainability of green innovation, revealing their synergistic relationship from a supply chain perspective. Methodologically, it provides a more refined empirical pathway for quantifying the sustainability of green innovation by segmenting the “capability–motivation” dimension and introducing moderating variables. Practically, it provides policymakers and corporate managers with empirical evidence regarding key moderating factors in the digital transformation process, thereby optimizing resource allocation and implementation pathways for green innovation strategies.

2. Literature Review

2.1. Cross-Supply Chain Spillovers from Digital Transformation

The business activities of enterprises rely on upstream and downstream financial transactions and supply and demand relationships to form a supply chain connection, customers and suppliers in the dynamic relationship of interdependence, as the market division of labor evolves from transactional buyer–seller relationships to collaborative partnerships [10,11], which prompted academics to expand the digital transformation research perspective to the supply chain interconnectedness and ripple effects. Informed by the social network theory, Hertzel et al. [12] propose that supply chain actors construct a multidimensional innovation nexus by integrating logistic operations, financial capital circulation, and data intelligence, positioning supply chains as pivotal innovation diffusion channels. Falcone et al. [13] propose that supplier firms leverage supply chain knowledge spillovers to influence customer innovation, while digital transformation reconfigures these innovation diffusion pathways [14]. Belhadi et al. [15] demonstrate that digital transformation empowers focal firms to mitigate carbon uncertainty by enhancing supply chain transparency, ultimately optimizing low-carbon performance through technology-enabled monitoring. Yan et al. [16] argue that supplier firms’ digital transformation using digital technologies creates new opportunities for corporate upgrading, organizational upgrading, and collaborative progress across the supply chain network. Liu et al. [17] reveal that digital transformation diminishes customer dependency risks by diversifying supply chain partnerships, while simultaneously reconfiguring focal firms’ structural roles within the network.

2.2. Digital Transformation as a Catalyst for Corporate Green Innovation

While substantial research exists on the digital transformation–green innovation nexus, empirical findings remain inconclusive, particularly regarding digital transformation’s catalytic role in driving sustainable corporate innovation. Empirical studies by scholars demonstrate that digital transformation exerts a statistically significant stimulating effect on green innovation outcomes among publicly traded companies [18,19], while others argue for an inverted U-shaped effect [20]. The root of these contradictory conclusions largely stems from the prevalence of empirical studies in related fields, which only summarize innovation performance in terms of aggregate indicators, while ignoring the differentiated connotations of different dimensions, such as green technological breakthroughs and the expansion of innovation scenarios. It is worth noting that digital technology has become the core driver of green innovation in Chinese enterprises as they accelerate the adoption of advanced digital technologies [21]. The multidimensional nature of green innovation mechanisms necessitates a paradigm shift in evaluation metrics, moving beyond traditional frameworks to resolve the theory–practice dichotomy.

At present, scholars investigate how digital transformation shapes green innovation trajectories through dual analytical lenses: internal and external drivers. At the internal driver level, existing research mainly focuses on the enabling mechanisms through which enterprises’ independent digital initiatives drive green innovation. From a process-oriented perspective, it explores how digital technologies reduce the costs and risks associated with green innovation by leveraging mechanisms such as resource orchestration and information transmission. For instance, Xue et al. [22] and Xu et al. [23] conducted econometric analyses on panel data of Chinese enterprises, revealing that digitalization optimizes human capital, boosts media attention, and eases financing constraints; they form synergies that drive progress in sustainable technology development. In terms of corporate governance, Liu et al. [24] confirm that digital upgrading strengthens the sustainability of green innovation by adjusting compensation incentives; Zhao et al. [25] note that analysts’ attention orchestrates the digital–green innovation nexus; Feng [26] demonstrates, based on signaling theory, that digital disclosure mitigates information asymmetry; and Liu [27] emphasizes that a robust digital internal control system helps prevent innovation-related risks. At the external environment level, studies center on transmission paths and moderating effects. Feng et al. [28] find that R&D investment and government subsidies act as transmission channels between digital transformation and green innovation, while also pointing out that industry-specific institutional pressure weakens the positive correlation between the two. In contrast, Zhu et al. [29] verify that market-driven environmental pressure exerts a positive moderating effect. Drawing on structural weighting theory, He [30] reveals that regulatory pressure and international opportunities enhance the aforementioned mediating effect. Taking construction enterprises as a case study, Li et al. [31] highlight the synergistic moderating role of external contextual systems, including policies, markets, and digital culture. Collectively, these studies offer systematic theoretical support for understanding how digital transformation drives enterprises’ green innovation efforts.

While prior research has advanced our knowledge of how digital transformation influences green innovation within enterprises, significant limitations persist in current academic investigations. On one hand, studies have revealed micro-level pathways through which digital technologies influence green innovation within individual enterprises, yet they exhibit clear limitations: first, they overly rely on aggregate innovation metrics such as “green patent counts”, which fail to seize the dynamic, continuous nature of green innovation throughout the process from investment to implementation; second, they tend to neglect the extended temporal dimension inherent in sustainable innovation processes. Sustainable innovation exhibits distinct temporal and financial characteristics when contrasted with conventional technological advancement, including extended maturation periods, elevated expenditure requirements, and heightened outcome variability [32]. Its “sustainability” holds far greater research value than merely whether it “occurs or not”, yet existing findings pay insufficient attention to this aspect. Furthermore, contemporary research underestimates the networked externalities generated when digital pioneers trigger green innovation cascades across vertical supply chain hierarchies, failing to reveal the cross-entity transmission mechanisms of digital capabilities within supply chains. Additionally, insufficient research exists on how supply chain “interdependence” moderates digital spillover effects, making it difficult to explain the differentiated causes of green innovation performance across varying supply chain structures.

Addressing these limitations, this paper adopts a supply chain perspective grounded in dynamic capability theory and institutional theory. By investigating the causal linkage between downstream digital transformation and upstream green innovation sustainability, it reveals the underlying mechanism. This study addresses the theoretical void concerning how digital supply chain transformation dynamically sustains environmental innovation, thereby providing theoretical and practical guidance for enterprises seeking long-term green innovation development through digital transformation within supply chains.

3. Theoretical Analysis and Hypothesis Formulation

3.1. Dual Mechanisms for the Sustainability of Green Innovation

Green innovation manifests as the embedding of green concepts into integrated design frameworks through hardware or software technology innovation to achieve energy efficiency, cleanliness, waste utilization, and environmental management [33]. Sustainable green innovation constitutes a strategic embedding process; it originates from green innovation, enabling enterprises to align technological capabilities with intergenerational sustainability objectives [34].

The sustainability of corporate green innovation manifests through dual dimensions: (1) the enduring capacity to sustain eco-innovation processes and (2) the driving force to perpetuate green innovation initiatives. Sustainability is the ability of enterprises to absorb and transform green technologies and experiences across organizations, to form a cross-cycle and stable iterative R&D and process optimization mechanism, to avoid the risk of innovation disruption, to carry out long-term and stable environmentally oriented innovation activities, and to respond to environmental shifts, the core of which is embodied in the knowledge-driven continuity of resource integration and the dynamic stability of the market response [35]. Continuous motivation is the internal willingness and external thrust that incentivizes sustained corporate engagement in green innovation, including the internal pursuit of a sustainable competitive advantage and environmental responsibility fulfillment [36], as well as external policy incentives [37], the market demand for green and supply chain synergistic pressure, etc., which interact with each other to help enterprises overcome the long cycle of green innovation, high costs, uncertainty, and other barriers and continuously invest in driving innovation activities. This study investigates how digital transformation influences both the sustaining ability and sustaining power of green innovation among supply chain enterprises, while unraveling its heterogeneous mechanisms of influence on the sustainability of green innovation. Figure 1 delineates the conceptual framework.

3.2. Customer Digital Transformation and Supplier Green Innovation Sustainability

The complementary lenses of resource dependence and transaction cost theories reveal multidimensional impacts of digital transformation on sustainable innovation [38]. Building upon this, this paper further integrates knowledge-based dynamic capability theory with institutional theory to construct a dual-mechanism framework of “capability–motivation”. The analytical framework reveals the continuous empowerment mechanism from downstream digital transformation to upstream sustainable innovation.

Capability mechanisms manifest in the processes of digital empowerment and dynamic capability upgrading. Dynamic capability theory highlights organizations’ capacity to continuously integrate, build, and restructure resources to respond to environmental shifts [8]. When combined with transaction cost theory, this framework empirically validates the sustainability mechanisms of green innovation. The digital transformation of customer substantially improves the demand transmission accuracy through advanced technological tools like big data analytics and intelligent platforms. This digital empowerment is fundamentally a process of knowledge sharing and capability restructuring. On one hand, real-time feedback on green standards and monitoring data from digital tools provide suppliers with a clear direction for resource integration, enabling them to allocate technical, financial, and other factors more efficiently to meet innovation demands. On the other hand, continuous digital oversight compels suppliers to establish rapid response mechanisms. This dynamic optimization of resources and processes ultimately ensures the sustainability of green innovation.

When viewed from the angle of motivation, the customer’s digital transformation strengthens its resource control and institutional transmission capabilities, forming a deep-seated driving force for supplier green innovation. Suppliers’ survival and development are highly dependent on key resources provided by customers, such as orders, funding, and market channels, granting customers significant power advantages within the supply chain [39]. When customers advance their digital transformation, they leverage big data and intelligent platforms to precisely communicate green requirements and convert them into mandatory standards. Under this coercive pressure stemming from resource control requirements, suppliers are compelled to integrate green innovation into their strategic core to maintain resource access. Simultaneously, within the digital environment, clients’ real-time sharing of high-quality suppliers’ green practices creates imitation pressure among other enterprises in the supply chain. Resource dependency further amplifies this imitation motivation, driving suppliers to proactively benchmark against leading players’ green innovation behaviors to avoid elimination in resource competition.

To summarize, green innovation sustainability focuses on enterprises’ integration of green innovation concepts and capabilities into long-term development strategies, which is mainly reflected in the dimensions of sustaining capability and sustaining power. Theoretical analysis reveals that the customer digital transformation operates through dual mechanisms to sustain supplier green innovation, thereby yielding the following research hypothesis:

H1. 

Digital transformation can promote the sustainability of green innovation, which manifests as enhancing sustained capability and strengthening sustained motivation.

3.3. Mechanism of Customer Digital Transformation Affecting Supplier Green Innovation Sustainability

3.3.1. Mechanism of Customers’ Digital Transformation on Suppliers’ Sustainable Green Innovation Capabilities

Digitalization dismantles informational barriers and reshapes industrial networks, thereby creating novel pathways for sustaining green innovative practices. Among them, knowledge spillover and the efficiency of supply and demand collaboration constitute key mediating variables. On the one hand, they promote the upgrading of suppliers’ innovation achievements; on the other hand, they reduce information asymmetry and resource misallocation. Together, they shape the trajectory of digital transformation, enhance the sustainability of green innovation, and accelerate the iteration of green technologies, strengthening the continuous driving force of enterprises for green innovation.

According to the knowledge base perspective, an enterprise’s competitive advantage and innovation ability are rooted in the knowledge resources it possesses [40]. The digital transformation among downstream customers constitutes a procedure of knowledge reconstruction and upgrading centered on data, technology, and processes [41], which conveys knowledge related to green innovation to upstream suppliers and thereby systematically enhances their sustainable green innovation capabilities. This tacit and explicit knowledge exhibits natural spillover effects in collaborative interactions with suppliers [42], yet effective knowledge utilization hinges on the firm’s absorptive capacity. By absorbing, assimilating, and applying this external knowledge, suppliers can effectively address their capability gaps in green technology R&D and environmental management systems, thereby strengthening their internal knowledge foundation and capability reserves for green innovation. Cohen and Levinthal [43] specifically highlight that absorptive capacity exhibits self-reinforcing characteristics—innovative practices feed back into existing knowledge reserves, forming a positive feedback loop of “knowledge accumulation–innovation output–new knowledge.” This cycle reduces the technical complexity and management costs associated with sustained green innovation, enabling suppliers to maintain the continuity and longevity of such activities even under resource constraints. It fosters a stable, cross-cycle mechanism for iterative R&D and process optimization, ultimately achieving long-term, environmentally oriented innovation.

Supply–demand synergy efficiency consolidates the sustainability of green innovation from the market responsiveness dimension, ensuring the stability of continuous capabilities by optimizing the adaptability of digital transformation to resource allocation and risk response. The real-time demand mapping system constructed by digital twin technology enables supply chain enterprises to dynamically adjust the direction of green innovation. For instance, when the environmental protection preferences at the consumer end are immediately fed back to the production end through the blockchain traceability system, enterprises can continuously optimize green processes without worrying about the disconnection of innovative achievements from the market. In addition, digital platforms can alleviate the bullwhip effect through real-time demand synchronization by minimizing distortions in demand signals [44]. When the market demand fluctuates, the upstream and downstream share green order data in real time through digital tools and supply chain data. The upstream can adjust the pace of the green technology research and development according to the actual demand, while the downstream can stably feedback the innovation effect. This dynamic adaptability avoids the waste or interruption of innovation resources caused by demand distortion, ensuring that green innovation maintains a stable pace in the long term.

In conclusion, knowledge spillover mechanisms amplify the penetration intensity of digital transformation via cognitive diffusion channels, and optimized supply–demand coordination elevates the execution precision of digital initiatives through resource allocation optimization, thereby yielding the following research hypotheses:

H2a: 

Digital transformation enhances the sustained capability of green innovation by unleashing knowledge spillovers.

H2b: 

Digital transformation enhances the sustained capability of green innovation by improving the efficiency of supply–demand coordination.

3.3.2. Mechanism of Customers’ Digital Transformation on Suppliers’ Sustainable Green Innovation Motivation

Digital transformation, with the support of the mutual competition among enterprises and financial incentives, stimulates suppliers’ willingness to expand funding in green resources, fully unleashing the energy accumulation function of suppliers’ green innovation and strengthening enterprises’ sustained momentum.

Digitalization has restructured the competitive rules of the supply chain, making green innovation a hard threshold for the survival of enterprises. This has forced upstream suppliers to continuously increase their investment in green R&D to enhance their competitiveness [45], which in turn has been transformed into a sustainable driving force. The digital upgrade of customers will break down information barriers, allowing upstream enterprises to clearly perceive the changes in the competitive landscape. If upstream and downstream enterprises fail to meet the green standards set by focus enterprises through digital tools, they may face the risk of being replaced. This pressure compels market actors to transition from compliance-driven approaches to strategic investments in sustainable technologies. Meanwhile, digital tools have accelerated the diffusion of green innovation information in the industry. Upstream and downstream enterprises can observe the technological progress of their peers in real time through digital platforms. This benchmarking pressure will strengthen their willingness to innovate and not fall behind and drive enterprises to continuously increase R&D investment to preserve market leadership. In addition, the supply chain visibility brought about by digital transformation will amplify the influence of market competition. The preferences of end consumers for green products can be quickly transmitted to the upstream through digital channels, putting enterprises that do not participate in green innovation at risk of shrinking market shares and further consolidating their motivation for long-term investment.

Tax incentives provide a guarantee for sustained momentum from the perspectives of costs and benefits, stabilizing enterprises’ willingness to invest by lowering the threshold for innovation and clarifying the expected benefits [46]. As organizations undergo a digital metamorphosis, workforce realignment commonly follows. Their digital platforms can accurately identify the green innovation demands of upstream enterprises, achieve a precise allocation of financial incentives, directly reduce the initial investment costs of enterprises, and alleviate the decline in motivation caused by high risks of innovation. Digital tools can bind financial incentives with innovation achievements. Focus enterprises can automatically realize the purchase premium of green products through smart contracts or ensure that subsidy funds are only used for environmental protection technology research and development through blockchain traceability systems. The certainty of such returns will strengthen enterprises’ expectations that green innovation is “profitable”, promoting short-term investments to be transformed into long-term persistence. Digital transformation can also help upstream and downstream enterprises more efficiently connect with external financial resources, shorten the realization cycle of policy dividends, enable enterprises to continuously feel the actual benefits produced by innovation, and thereby consolidate their impetus for green innovation.

In conclusion, the competitive effect, by establishing a pressure mechanism of “no innovation means elimination”, mandates continued corporate investment in sustainable R&D. Fiscal incentives form an incentive mechanism of “innovation is beneficial” by reducing costs and clarifying benefits, guiding enterprises to proactively continue their innovative behaviors, thereby yielding the following research hypotheses:

H3a: 

Digital transformation enhances the sustained motivation for green innovation by intensifying market competition.

H3b: 

Digital transformation enhances the sustained motivation for green innovation by increasing fiscal incentives.

3.4. Analysis of Regulatory Effect

Supply chain centralization demonstrates the extent of the interdependence between customers and suppliers and to a large extent determines the diffusion of digitalization across supply chain echelons and fundamentally reshapes green innovative trajectories.

The supplier market concentration fundamentally moderates buyer-side digitalization’s impact on sustainable production through dependency dynamics [47]. Under a high supplier concentration, core customers’ digital initiatives impose stronger constraints and incentives on suppliers. To maintain strategic partnerships, suppliers must closely align with their clients’ ecological demands, leading to a sustained investment in clean-tech R&D that ultimately catalyzes industrial ecosystem upgrading. Conversely, low-concentration scenarios weaken single-customer dependence, diminishing digitalization’s coercive power over suppliers’ environmental strategies. Such firms are more inclined to self-calibrate their innovation trajectories according to endogenous capabilities and competitive landscapes, thereby diluting the digitalization-driven synchronization effect. On the other hand, the customer concentration regulates the above relationship through the “input–synergy” mechanism [48]: when the customer concentration is high, customers have a stronger motivation to deeply navigate the ontogenesis of suppliers to ensure that the green output of suppliers can stably meet the sustainable needs of their own supply chains. This collaborative behavior will amplify the beneficial influence of customers’ digital transformation on the sustainability of suppliers’ green innovation. When the customer concentration is low, the attention and resource investment of customers towards specific suppliers decrease. Their digital transformation focuses more on improving their own operational efficiency, which causes the guidance and support for the green innovation of suppliers to weaken, thereby yielding the following research hypotheses:

H4a: 

The supplier concentration positively moderates the effect of digital transformation on green innovation sustainability.

H4b: 

The customer concentration positively moderates the effect of digital transformation on green innovation sustainability.

Table 1. Summary of hypotheses.

Symbol	Hypothesis
H1	Digital transformation can promote the sustainability of green innovation, which manifests as enhancing sustained capability and strengthening sustained motivation.
H2a	Digital transformation enhances sustained capability of green innovation by unleashing knowledge spillovers.
H2b	Digital transformation enhances sustained capability of green innovation by improving the efficiency of supply–demand coordination.
H3a	Digital transformation enhances sustained motivation of green innovation by intensifying market competition.
H3b	Digital transformation enhances sustained motivation of green innovation by increasing fiscal incentives.
H4a	Supplier concentration positively moderates the effect of digital transformation on green innovation sustainability.
H4b	Customer concentration positively moderates the effect of digital transformation on green innovation sustainability.

summarizes all hypotheses proposed in this study, clearly presenting the core content and logical relationships of each hypothesis.

4. Study Design

4.1. Sample and Data

China’s State Council promulgated the 12th Five-Year Development Agenda (2011–2015), a comprehensive blueprint guiding socioeconomic progress. The policy mandate to accelerate the convergence between informatization and industrial modernization laid the institutional foundation for corporate digital upgrading. Since then, China’s technological ecosystem evolution has accelerated into a phase of mass adoption and exponential growth. Therefore, this article sets the research period from 2011 to 2023. We target publicly traded firms on China’s A-share exchanges as the analytical sample. Since the upstream supplier (S) may correspond to multiple downstream customers (X, Y, Z) in a certain year (such as 2011), as documented by Isaksson et al. [49], we utilized supply chain information data of publicly traded companies within China’s economy–finance research datasets. We constructed the observed values of S-X-2011, S-Y-2011, and S-Z-2011 and matched the supply and sales list of listed companies with the data on enterprise financial indicators and corporate governance in this database according to the stock code and year. Other raw data mainly came from the China Statistical Yearbook, Guotai ‘an Economic and Financial Research Database (CSMAR), China Research Data Service Platform (CNRDS), Wind-sourced market metrics, and corporately disclosed annual filings. At the same time, drawing on existing research practices, the original data was handled as follows: financial enterprises, ST enterprises, *ST enterprises, PT enterprises, and those with severe data deficiencies are excluded. To mitigate outlier-induced bias, all continuous predictors underwent bilateral 1% winsorization at the distribution tails, and finally 2002 observations were obtained.

4.2. Variable Measurement and Description

4.2.1. Explained Variables

The continuous capacity (GI_C) and continuous motivation (GI_M) of suppliers for green innovation, the core of maintaining an enterprise’s sustainable capacity for green innovation, lie in the comprehensive strength to stably produce green innovation achievements over the long term. Green patents are innovative knowledge assets certified by authoritative institutions [50,51], and the average annual growth of their application volume not only reflects the cumulative breakthroughs of enterprises under the framework of green transition policies but also demonstrates the sustainability and quality stability of innovation activities. Consequently, we adopt the annual growth rate of green-focused patent filings as a proxy for continuous capacity. The sustained motivation in business entities lies in their willingness and ability to maintain persistent capital allocation. The key is to accumulate an impetus for green innovation through the continuous injection of resources. Green R&D investment encompasses three core pillars: financial capital allocation, talent pool development, and technological infrastructure. The sustained growth trajectory of this investment reflects enterprises’ institutionalized commitment to green innovation, where incremental resource inputs create a compounding accumulation effect akin to a rolling snowball. Therefore, this study employs the growth rate of green R&D expenditure to measure the continuous motivation of green innovation. At present, enterprises have not separately disclosed data on green R&D investment. So, we adopt an indirect calculation method of coupling the total R&D investment with the share of green patents [52,53]. The specific approach is as follows: This study calculates the green patent ratio by dividing the annual number of enterprise green patent authorizations by the total patent grants, thereby approximately measuring the scale of green R&D investment. Among them, if an enterprise has green patents, the total R&D investment can be approximately split according to the ratio of “the number of green patents/the total number of patents” [54]. If an enterprise belongs to the green industry, it can be considered that all its R&D investments are green R&D [55]. Finally, we calculate the annualized green R&D expenditure growth coefficient to capture the stability of the enterprise’s resource inclination towards green innovation over a longer period.

4.2.2. Explanatory Variables

Customer digital transformation (DT). During China’s economic transition period, enterprise digitalization has emerged as a core metric for evaluating high-quality development, attracting significant attention from both academia and industry practitioners. The early research adopted a single dimension of measurement indicators, which made it difficult to represent the enterprise-wide digital integration of business processes. With the advancement of machine learning techniques and the increasing accessibility of textual big data, many scholars have created enterprise digital terminology databases [56,57]. This study follows the methodological framework established by Zhai et al. [58], with corporate annual reports serving as the primary textual data source. Through text analysis, it measures the incidence of “digital transformation-related keywords” to assess DT. The rationale for selecting annual reports as the analytical vehicle lies in their lexical usage, reflecting corporate strategic characteristics and future planning. The high frequency of keywords often aligns intrinsically with companies’ actual investments and strategic deployments in the digital domain. The implementation process comprises the following stages: Digital transformation is divided into nine dimensions, namely machine intelligence, large-scale data analytics, distributed computing and edge computing, Distributed Ledger Technology (DLT), Internet of Things and industrial Internet, emerging technologies, digital management, digital production, and diverse Internet ecosystems, and 118 indicators are constructed (The keyword dictionary and frequency distribution is included in Table S1). Then, Python 3.13 is used for word segmentation to count the total number of words for each indicator. To reduce statistical bias, the sections of symbols, numbers, and English letters, as well as the keyword statistics sections indicating negative words before and after the keyword fields, are excluded. Due to the obvious right-skewed feature of word frequency data, it is necessary to perform logarithmic processing.

4.2.3. Control Variables

For result robustness, we build upon the findings of Tao et al. [59] and Cheng et al. [60], and considering that a company’s own financial health and industry-specific regulatory measures significantly impact green innovation, we select the firm tenure (Age), company scale (Size), total asset turnover rate (TAT), intensity of financing constraint (SA), proportion of independent directors (NID), cash flow ratio (CF), resource allocation efficiency (TQ), and environmental regulation (ER) as control variables. Firm tenure (Age) measures the time elapsed since a firm’s market debut. Mature enterprises typically possess more stable resource reserves and industry experience, potentially enabling a greater capacity to sustain long-term green innovation investments. The company scale (Size), derived from applying logarithmic transformation to the balance sheet’s total asset figure, controls for size effects. Larger enterprises often demonstrate stronger financial and technological capabilities to support the sustained development of green innovation, while also benefiting from economies of scale to spread innovation costs. The total asset turnover rate (TAT) is derived by dividing the fiscal year operating revenue by the closing balance of total assets, reflecting a company’s asset operational efficiency. Efficient asset turnover enhances resource utilization and provides sustained capital recycling support for green innovation. The financing constraint intensity (SA) uses the SA index. Higher financing constraints restrict firms’ resource allocation toward innovation, consequently undermining the long-term viability of green innovation. The independent director ratio (NID) computes the ratio of independent directors to the entire board membership. It typically provides more objective oversight and strategic advice. Increasing their proportion may strengthen corporate commitment to sustainable development and promote long-term planning for green innovation. The cash flow ratio (CF) equals the operating cash flow to total assets ratio, reflecting the sufficiency of a firm’s internal funds. A stable cash flow is a critical foundation for sustaining green innovation investments, reducing the risk of the innovation disruption caused by short-term funding shortages. Resource allocation efficiency (TQ) is measured using Tobin’s Q ratio. An elevated Tobin’s Q signals the market’s perception of profitable investment opportunities, facilitating resource allocation toward high-efficiency domains, including green technology R&D and sustainable innovation enhancement. Environmental regulation (ER) intensity is operationalized as the proportion of pollution abatement expenditures to the total industrial output value, measured at the two-digit industry classification level. Incorporating this variable effectively controls for potential interference from environmental policies that may influence corporate environmental behavior and thereby affect the sustainability of green innovation. In addition, this article also takes into account the influence of the year and industry.

Table 2. Variable definitions and measurements.

Type	Variable Name	Symbol	Measurement
Explained Variable	Continuous Capacity	GI_C	The growth rate of green patent applications.
	Continuous Drive	GI_M	The growth rate of green R&D investment.
Explanatory Variable	Digital Transformation	DT	An index constructed by extracting keyword frequencies through text analysis.
Control Variable	Firm Tenure	Age	The time elapsed since a firm’s market debut.
	Company Scale	Size	Applying logarithmic transformation to the balance sheet’s total asset figure.
	Total Asset Turnover Rate	TAT	Operating Income/total Assets at the end of the period.
	Intensity of Financing Constraint	SA	The SA index.
	Proportion of Independent Directors	NID	Number of independent directors/board size.
	Cash Flow Ratio	CF	Net cash flow/total assets,
	Resource Allocation Efficiency	TQ	The Tobin Q index.
	Environmental Regulation	ER	The proportion of pollution abatement expenditures to total industrial output value.

elaborates on the definitions and specific measurement methods for each variable, categorized by type, and also provides corresponding symbols for them.

4.3. Model Construction

The Hausman test decisively considers the fixed effects model as more appropriate than the random-effects model specification; we employ the fixed effects estimator to examine (i) the heterogeneous effects of digital transformation on green innovation sustainability and (ii) their asymmetric relationship pathways. Model (1) is used for testing:

$${\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{D}\mathrm{T}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\theta }}_{\mathbf{n}}{\mathbf{X}}_{\mathbf{i},\mathbf{t}}^{\mathbf{n}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(1)

with i as the cross-sectional unit (firm) and t as the time dimension (year); ${\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{t}}$ is the explained variable, including the enterprise’s continuous capacity for green innovation (GI_C) and continuous motivation (GI_M); ${\mathrm{D}\mathrm{T}}_{\mathrm{i},\mathrm{t}}$ serves as the primary independent variable, quantifying enterprise-level digitalization intensity; ${\mathsf{\mu }}_{\mathrm{i}}$ represents the industry fixed effect; ${\mathsf{\delta }}_{\mathrm{t}}$ represents the time-invariant fixed effects (year dummies); and the stochastic disturbance term ${\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$ follows a normal distribution. Given the multidimensional determinants shaping corporate green innovation, we simultaneously introduced a set of control variables ${\mathbf{X}}_{\mathbf{i},\mathbf{t}}$.

5. Empirical Analysis

5.1. Descriptive Statistics

The descriptive analysis of primary variables is documented in

Table 3. Distribution characteristics of key variables.

Variable	N	Mean	Min	Max	SD
GI_C	2002	9.081	−6.993	38.395	4.306
GI_M	2002	3.610	−9.184	7.659	5.135
DT	2002	2.427	0.693	6.757	1.233
Age	2002	13.225	1	33	7.051
Size	2002	22.982	18.291	31.431	1.601
TAT	2002	0.640	0.001	7.571	0.438
SA	2002	−3.788	−4.836	−0.288	0.361
NID	2002	0.374	0	0.8	0.054
CF	2002	0.047	−0.650	0.533	0.065
TQ	2002	1.772	0.611	27.338	1.086
ER	2002	2.285	0.782	5.196	1.563

.

The data in the table shows that the average growth rate of the stock of green patents (GI_C) is 9.081, indicating that enterprises have a certain sustainable ability for green innovation and development. However, its standard deviation is 4.306, suggesting that there are significant differences in the sustainable ability of green innovation among enterprises, and the distribution span is large. The average growth rate (GI_M) of enterprises’ green R&D investment is 3.610, which is lower than the average of the sustainability capacity, where the measure of spread equals 5.135. The findings reveal a moderately inadequate momentum for sustainable technological innovation at the organizational level, and the degree of the dispersion of the driving force level is higher, with an uneven performance among enterprises. Therefore, a deeper investigation into the drivers of these differences and their varying levels is warranted in order to help enterprises enhance their continuous performance in green innovation. The mean score for the digital transformation attainment is 2.427, and the data exhibits substantial dispersion, indicating a polarization in the digitalization levels of enterprises. Leading enterprises have achieved deep transformation, while those at the bottom are still in the basic stage. In conclusion, exploring the differentiated mechanism of digital transformation on the sustainability of enterprises’ green innovation and its boundary conditions carries both profound theoretical implications and pressing practical relevance. The data shows no obvious bias, indicating that the processing is relatively reasonable. The operationalization of control variables mirrors extant theoretical specifications.

5.2. Baseline Regression

Table 4. Baseline estimation tests.

Variable	GI_C				GI_M
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
DT	0.034 ***	0.023 ***			0.091 ***	0.061 ***
	(0.011)	(0.011)			(0.012)	(0.012)
L1. DT			0.025 ***				0.047***
			(0.010)				(0.012)
L2. DT				0.012 ***				0.039 ***
				(0.010)				(0.012)
Constant	2.470 ***	1.757 ***	0.273 ***	0.555 ***	1.730 ***	4.193 ***	2.950 ***	2.606 ***
	(0.045)	(0.444)	(0.218)	(0.356)	(0.033)	(0.459)	(0.525)	(0.634)
Control	No	Yes	Yes	Yes	No	Yes	Yes	Yes
Id FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Hausman chi2	78.68 ***	53.24 ***			48.26 ***	157.08 ***
N	2002	2002	2002	2002	2002	2002	2002	2002
R-sq	0.101	0.112	0.108	0.041	0.145	0.101	0.112	0.108

Note: Statistical significance is denoted as *** (p < 0.01), and robust deviations are reported within square brackets. The explained variables measurements were extracted from three authoritative financial databases (CSMAR, CNRDS, and Wind), while the explanatory variables originated from the annual reports of listed companies.

presents the Hausman test results, which prove the rationality of the model specification, and also shows the key regression estimates.

Univariate regression analyses (columns 1 and 5) reveal a statistically robust positive relationship (p < 0.01) between DT and dual dimensions of corporate green innovation—GI_C and GI_M—prior to the covariate adjustment. Per one-standard-deviation rise in the digital transformation level, the sustainability capability of corporate green innovation significantly increases by 0.034 standard deviations, while the driving force increases by 0.091 standard deviations. This preliminary finding suggests that digital transformation positively influences both dimensions. To mitigate confounding effects arising from firm-specific heterogeneity and exogenous environmental variations, our extended model incorporates comprehensive control variables encompassing the firm age, firm size, and total asset turnover. As evidenced in columns (2) and (6), this multivariate adjustment attenuates the estimated marginal effects of digital transformation at 0.023 and 0.061. This change indicates that certain confounding factors do indeed affect the strength of the association between the two, but the core conclusion remains unchanged. The robust positive causality between DT and persistent green innovation survives stringent econometric tests. Quantitatively, DT’s marginal effect on the green innovation impetus exceeds its impact on sustainability by a statistically significant margin. This result not only validates hypothesis H1 but also reveals that digital transformation exerts a stronger influence on GI_M than on GI_C. This may be related to the fact that digital tools are more likely to activate corporate innovation willingness through market signals. To mitigate potential endogeneity concerns arising from the simultaneity bias and contemporaneous omitted variables, and accounting for the endogenous selection pattern where environmentally proactive firms tend to accelerate digital adoption, our empirical strategy introduces first- and second-order lagged digital transformation indicators. The statistically consistent estimates (p < 0.05 across lagged terms) corroborate the primary regression outcomes.

5.3. Endogeneity Test

5.3.1. Reverse Causality Test

To mitigate endogeneity concerns arising from the interrelationship between firms’ digitalization levels and sustainable innovation activities, which may lead to endogeneity problems, the research design incorporates an IV framework to further mitigate the impact of this reverse causal relationship on the research results. Firstly, drawing on Lewbel’s [61] research, the instrumental variable (DT_IV1) is developed using the digital transformation intensity of peer enterprises within the same provincial administrative region. The justification for this variable is that digital transformation across enterprises in the same province is often shaped by common external factors like regional policies, thus showing a strong correlation with the firm’s own digital transformation level. However, the cubic term of this difference employs a nonlinear transformation to isolate the firm-level heterogeneity, significantly weakening its direct association with the firm’s green innovation persistence and thus satisfying the exogeneity requirement. Secondly, considering the validity of DT_IV1, this study constructs DT_IV2 by deriving the industry–year averaged digital transformation index from peer firms [62]. The instrument’s validity stems from industry-level homogeneity, where firms sharing the same sector face isomorphic technological trajectories and competitive landscapes, while the industry digitalization mean maintains strong predictive power for individual firms’ transformation levels. At the same time, this variable reflects the overall characteristics of the industry rather than the innovation decisions of individual enterprises themselves and thus has a weaker endogenous relationship with the sustainability of green innovation within the enterprise. The two together form the instrumental variable group. After numerous tests, the empirical analysis confirms the validity of our instrumental variables, with no evidence of weak identification or overidentification issues. As presented in

Table 5. Instrumental variable regression results.

Variable	GI_C		GI_M
	(1)	(2)	(3)	(4)
DT_IV1	0.024 ***		0.742 ***
	(0.020)		(0.340)
DT_IV2		0.018 ***		0.736 ***
		(0.147)		(0.308)
Control	Yes	Yes	Yes	Yes
Id FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
N	1822	1688	1822	1688
Kleibergen–Paap	692.424 ***	499.292 ***	692.424 ***	499.292 ***
rk LM statistic
Cragg–Donald	1799.183	1149.799	1799.183	1149.799
Wald F statistic
Hansen J	22.990	14.895	7.881	3.208

Note: Statistical significance is denoted as *** (p < 0.01), and robust deviations are reported within square brackets. All instrumental variables utilize a digital transformation index calculated based on publicly listed companies’ annual reports.

, the second-stage IV regression results—controlling for both the year and firm fixed effects—demonstrate statistically significant positive effects of the instruments on firms’ green innovation sustainability and motivation. These robust findings further validate our core hypothesis.

5.3.2. Excluding Sample Selection Bias

Considering that the research samples may have non-random selection phenomena, which could lead to bias in the estimation results, this paper adopts a dual strategy to correct the potential problems. On the one hand, utilizing Heckman’s selection model framework with dual estimation stages comprising both selection and outcome equations, systematic biases caused by the self-selection behavior of samples are identified and corrected. On the other hand, combined with the PSM-DID regression, based on covariate information, control samples with similar characteristics are matched for the treatment group samples. While balancing the differences in observable variables between groups, the interference of the sample selection bias on the estimation of causal effects is weakened. The two methods complement each other, effectively alleviating the endogeneity problem caused by the non-random sample selection from different perspectives and enhancing the reliability of the research conclusions and the accuracy of the causal inference.

Firstly, extending the methodological framework of Lan et al. [63], we implement Heckman’s two-stage procedure to correct for the selection bias. In the initial stage, the dummy variable of digital transformation (DT_Dum1) is set as the dependent variable. A binary indicator is created, where 1 represents above-median digital transformation and 0 denotes below-median levels annually. At the same time, considering that enterprise-level factors, including the firm age, size, asset turnover, financing constraints, board independence, cash flow position, resource allocation efficiency, and market concentration, may systematically influence digital transformation, the Probit regression framework enters these factors as dependent covariates. The inverse Mills ratio (IMR) derived from the first-stage Probit estimation was systematically integrated into the second-stage OLS specification. As demonstrated in

Table 6. Heckman two-step method and PSM-DID.

Variable	Heckman		PSM-DID
	GI_C	GI_M	GI_C	GI_M
	(1)	(2)	(3)	(4)
DT	0.091 ***	0.117 ***	0.069 **	0.132 ***
	(0.012)	(0.018)	(0.014)	(0.011)
IMR	0.865 *	0.299 *
	(0.533)	(0.535)
Control	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
N	2002	2002	1957	1957
R-sq	0.130	0.116	0.103	0.241

Note: Statistical significance is denoted as *** (p < 0.01), ** (p < 0.05), and * (p < 0.1); robust deviations are reported within square brackets. The virtual variables are set based on foundational information from the CSMAR, CNRDS, and Wind databases.

, the core findings remain robust to this correction for selection bias.

Secondly, we further construct the virtual variable of digital transformation (DT_Dum2) and represent it as follows: firms are classified into the treatment group (D = 1) when their digital assets’ value-added proportion exceeds the 75th percentile of intangible assets in the current year, with remaining firms as controls (D = 0). Using Model 1’s covariates as matching dimensions, we implement nearest neighbor propensity score matching for the analysis [64]. This approach not only screens out enterprise groups with more comprehensive digital adoption through a quantile division, providing a clear group definition for the subsequent causal effect assessment, but also enables the counting of matched control firms to exhibit greater statistical significance than that in the treatment group, thus allowing for more a precise screening of samples with smaller deviations during matching. In this paper, nearest neighbor matching, the PSM balance test, and the kernel density distribution are adopted. Following the application of inverse probability weighting, the distributional alignment across experimental conditions demonstrates a marked enhancement. The subsequent DID estimation using the matched sample reconfirms the statistical significance of key coefficients, as evidenced in columns (3)–(4) of Table 6, thereby validating our baseline findings.

5.3.3. Placebo Test

To mitigate potential confounding effects from latent variables, we conduct a placebo test based on randomly assigned values of explanatory variables. Without changing the control variables and the explained variables, the original assignment of digital transformation indicators is disrupted and randomly assigned, and then the benchmark model is re-estimated. Figure 2 depicts the distribution and statistical significance levels of the predictor’s regression coefficients. The figure demonstrates that randomly assigned explanatory variables yield coefficient means near zero, following an approximately normal distribution with mostly insignificant estimates. The regression results pass the placebo test, indicating that the benchmark conclusion does not result from unobservable variables.

5.4. Robustness Test

5.4.1. Replacement of Important Variables

To rigorously validate the consistency of our empirical findings, multiple complementary verification procedures were systematically implemented. Regarding the dependent variables, this study addresses the significant disparities in innovation thresholds and the legal validity between green patent applications and granted patents. We substitute application counts with granted green patent counts as the core metric. This methodological shift—transitioning from the application to the grant stage—strengthens the robustness of the conclusions (see column 1,

Table 7. Replacement of important variables.

Variable	GI_C_Adjusted	GI_M_Adjusted	GI_C	GI_M
	(1)	(2)	(3)	(4)
DT	0.735 **	0.813 *
	(0.347)	(0.510)
DT_adjusted			0.259 **	0.135 *
			(0.040)	(0.010)
Control	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
N	2002	2002	2002	2002
R-sq	0.098	0.532	0.088	0.023

Note: Statistical significance is denoted as ** (p < 0.05), and * (p < 0.1); robust deviations are reported within square brackets. The explained variables’ alternative measurements were extracted from three authoritative financial databases (CSMAR, CNRDS, and Wind), while the explanatory variables’ proxy indicators originated from the official China Statistical Yearbook.

). To capture the persistent dynamics of green innovation, we employ the green patent citation frequency as a proxy for R&D investment (see column 2, Table 7). Such an approach provides a more precise quantification of the long-term value creation in green technological innovation, overcoming the limitations of traditional R&D expenditure metrics in assessing innovation quality.

Among the independent variables, since organizational digitization constitutes a strategic initiative yielding substantial benefits, enterprises have the subjective motivation to strategically overestimate the degree of their transformation, which makes it difficult for a keyword capture based on text analysis to completely avoid measurement biases. We use the mean annual digital transition subsidy as the proxy variable for the further examination of digital transformation. As a policy incentive tool, government fiscal support intensity correlates positively with the organizational technological capability to obtain resources and carry out transformation activities, which can effectively eliminate the overestimation of digital transformation. The estimation findings appear in columns (3)–(4) of Table 7.

The re-estimated models under alternative specifications yield coefficients aligned with the baseline estimates, confirming the statistical reliability of our findings.

5.4.2. Eliminate Confounding Factors

As symbiotic elements of regional economic ecosystems, digital transformation and green innovation demonstrate mutually reinforcing interdependencies. They are not only constrained by external conditions such as the financial ecological environment, industrial policy orientation, and market supervision rules of the enterprise’s location but also interact with internal elements as seen in the business’s own resource endowment and governance structure, making the exploration of the intrinsic connection between the two face multiple interferences. To strip away the influence of confounding factors and accurately reveal their causal logic, this study implemented the following experimental protocols:

First, to ensure analytical consistency, cities exhibiting advanced digital transformation levels were excluded from the sample [65]. Regional disparities in digital development are well-documented, with the “China Urban Digital Economy Development Report 2022” identifying Beijing, Shanghai, Shenzhen, Guangzhou, and Hangzhou as the top five cities based on annual digital competitiveness metrics. These five cities have demonstrated significant leading advantages in areas such as the construction of an R&D and innovation paradigm, the aggregation of high-end elements, and policy coordination and guarantees. The efficiency of their digital-driven development is accelerating its transformation into the core driving force for enhancing the city’s capacity, and the technological transition produces more substantial measurable effects. Therefore, we excluded enterprises headquartered in these five cities (as shown in

Table 8. Regression results after controlling for confounding factors.

Variable	GI_C	GI_M	GI_C	GI_M	GI_C	GI_M
	(1)	(2)	(3)	(4)	(5)	(6)
DT	0.027 ***	0.079 **	0.020 **	0.070 ***	0.056 ***	0.109 ***
	(0.018)	(0.012)	(0.011)	(0.012)	(0.019)	(0.022)
Control	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
N	1430	1430	1586	1586	1885	1885
R-sq	0.081	0.219	0.082	0.243	0.098	0.233

Note: Statistical significance is denoted as *** (p < 0.01), ** (p < 0.05), and robust deviations are reported within square brackets.

, columns 1–2) and re-estimated the regression model.

Secondly, we eliminate samples from computer-related industries. As a core and cutting-edge field of digital transformation, the computer industry has distinct industry characteristics in terms of the pace of technological iteration, the construction of an innovation ecosystem, and the logic of market evolution. Its inherently highly digitalized industrial attribute may enable it to present unique patterns in its digital transformation path and green innovation model that are different from those of traditional industries. Therefore, this paper excludes three industry samples: “ICT manufacturing sectors”, “digital technology services”, and “web-based service industries”. The corresponding regression outcomes are displayed in columns (3)–(4) of Table 8.

Finally, enterprises with zero green patent citations and applications during the sample period are excluded. Some enterprises may not yet have carried out green innovation activities, and the inclusion of samples would interfere with the identification of the true causal relationship between variables. By eliminating such samples, we restrict our analysis to firm cohorts with verified green technology development activities, enhance the economic significance of the variables, and refine the assessment of how technologies enable green technology development. The regression analysis is repeated in columns (5) and (6) of Table 8, confirming the initial findings.

After reconstructing the sample through tripartite robustness checks and performing renewed econometric estimations, the outputs presented in Table 8 consistently demonstrate that corporate digitalization continues to exert statistically significant positive effects on the persistence of eco-innovation activities. These findings confirm the robustness of our core thesis against potential confounding variables.

5.4.3. Employing an Alternative Regression Method

The sensitivity of the model setting exerts a direct causal effect on the reliability of the key takeaways. Owing to the differences among various econometric models in terms of the variable relationship setting, error term distribution assumptions, and parameter estimation methods, significant variations in estimation results may occur. To enhance the credibility of our findings, we perform robustness analyses to validate our findings by employing alternative econometric specifications. While preserving the operational definitions of core covariates (independent/dependent variables and controls), the Tobit regression framework is introduced to address potential censoring characteristics of the data. As evidenced in

Table 9. Tobit regression results.

Variable	GI_C	GI_M
	(1)	(2)
DT	0.123 ***	0.172 ***
	(0.012)	(0.016)
Control	Yes	Yes
Firm FE	Yes	Yes
Year FE	Yes	Yes
N	2002	2002
R-sq	0.052	0.089

Note: Statistical significance is denoted as *** (p < 0.01), and robust deviations are reported within square brackets.

of the regression analysis, the coefficient signs and significance levels remain materially unchanged from our baseline estimates.

5.5. Heterogeneity Analysis

5.5.1. Nature of Supplier Ownership

From the perspective of the heterogeneity of suppliers’ property rights, there are significant differences in the innovation-driven mechanisms between non-state-owned and state-owned enterprises, as shown in

Table 10. Analysis of heterogeneity in nature of supplier ownership.

Variable	GI_C		GI_M
	State-Owned	Non-State-Owned	State-Owned	Non-State-Owned
	(1)	(2)	(3)	(4)
DT	0.100 ***	0.168 ***	0.184 ***	0.127 ***
	(0.412)	(0.009)	(0.421)	(0.011)
Constant	2.471 ***	1.695 ***	8.036 **	3.148 ***
	(3.684)	(0.365)	(1.334)	(0.415)
Control	Yes	Yes	Yes	Yes
Id FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
N	568	1434	568	1434
R-sq	0.289	0.040	0. 457	0.186

Note: Statistical significance is denoted as *** (p < 0.01), ** (p < 0.05), and robust deviations are reported within square brackets.

.

For non-state-owned enterprises (columns (2) and (4)), the coefficient of DT on GI_C is larger than GI_M. Due to resource constraints in external financing and policy support, non-state-owned enterprises are more dependent on external resources, such as customer digital transformation, to enhance their innovation capabilities. The lack of such resource acquisition strengthens their willingness to break through innovation bottlenecks through external cooperation. Their green innovation stems more from the need to improve efficiency under environmental protection pressure. This market makes sustaining innovation a crucial way to maintain competitiveness. Thus, customer digital transformation, by providing a stable external impetus for knowledge and resource acquisition, has a more pronounced impact on ensuring the continuity of their green innovation efforts.

For state-owned enterprises (columns (1) and (3)), the coefficient of DT on both GI_C and GI_M is significantly positive, and the driving effect of customer digital transformation on GI_M is more pronounced than that on GI_C. Strategy-driven innovation often prioritizes achieving large-scale, impactful breakthroughs to align with macro-policy requirements. Customer digital transformation provides state-owned enterprises with targeted resource support that matches these strategic demands, thereby exerting a stronger impetus on the motivation of green innovation. In contrast, their innovation continuity is more secured by stable institutional guarantees and long-term policy support rather than external digital empowerment, resulting in a relatively weaker driving effect of DT on GI_C compared to GI_M.

5.5.2. Customer Market Influence

To explore the heterogeneity of the customer market influence, we divided the sample into two groups based on the proportion of customer sales, those with a stronger downstream customer market influence and those with a weaker influence, and conducted a regression analysis. As presented in

Table 11. Analysis of heterogeneity in customer market influence.

Variable	GI_C		GI_M
	Strong	Weak	Strong	Weak
	(1)	(2)	(3)	(4)
DT	0.089 ***	0.078 ***	0.184 ***	0.118 ***
	(0.013)	(0.012)	(0.421)	(0.011)
Constant	1.383 ***	2.109 ***	8.036 **	3.537 ***
	(0.576)	(0.454)	(1.334)	(0.519)
Control	Yes	Yes	Yes	Yes
Id FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
N	966	1036	966	1036
R-sq	0.040	0.045	0. 457	0.165

Note: Statistical significance is denoted as *** (p < 0.01), ** (p < 0.05), and robust deviations are reported within square brackets.

.

The results reveal that downstream customers’ market position significantly affects the transmission of suppliers’ innovation effects, and specifically, the digital transformation of customers with greater market influence has a more obvious promoting effect on the green innovation continuity of upstream suppliers. It can be attributed to how high-influence customers act on the core facets of innovation persistence. Firstly, these customers, via supply chain collaboration and technical standard dissemination, transfer systematic and advanced knowledge. This knowledge does not just offer immediate solutions to suppliers’ technical bottlenecks (fueling short-term innovation motivation) but also gets internalized into suppliers’ institutionalized knowledge management systems and operational routines. Over time, this helps build the long-term capability for sustained green innovation (GI_C). Secondly, the resource support they provide, such as funds, cutting-edge technologies, and market access, is not limited to project-specific, short-term goals. Instead, it often comes with long-term collaboration commitments, enabling suppliers to make continuous investments in R&D infrastructure and talent development, which are crucial for maintaining green innovation over an extended period. Thirdly, the strict innovation requirements from high-influence customers create a “reverse driving mechanism” that pushes suppliers to not only increase the short-term innovation input but also establish internal processes for continuous improvement and optimization. These processes, like regular R&D reviews and talent training programs, are essential for the enduring capability of green innovation. In contrast, for customers with weak market influence, due to insufficient knowledge and resource input, their digital transformation fails to bring about such comprehensive and lasting impacts on suppliers’ innovation systems, resulting in a less significant boost to green innovation continuity. All the above findings regarding the heterogeneity of customer market influence are detailed and supported in Table 11.

5.5.3. Regional Development Disparities

From the perspective of suppliers’ regional heterogeneity, we analyze the differential effects by combining regional development endowments and institutional environment differences, and the results are shown in

Table 12. Analysis of heterogeneity in regional development.

Variable	GI_C			GI_M
	Eastern	Central	Western	Eastern	Central	Western
	(1)	(2)	(3)	(4)	(5)	(6)
DT	0.180 ***	0.009 *	0.168 ***	0.285 ***	0.023 ***	0.260 ***
	(0.191)	(0.053)	(0.173)	(0.245)	(0.061)	(1.009)
Constant	5.258 ***	1.276	4.374	2.727 ***	3.535 ***	5.469 *
	(8.790)	(2.403)	(2.875)	(1.301)	(2.777)	(7.851)
Control	Yes	Yes	Yes	Yes	Yes	Yes
Id FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
N	861	612	529	861	612	529
R-sq	0.564	0.124	0.579	0.638	0.272	0.810

Note: Statistical significance is denoted as *** (p < 0.01), * (p < 0.1), and robust deviations are reported within square brackets.

.

For the eastern region (columns (1) and (4)), the coefficient of DT on GI_C and GI_M is 0.180 and 0.285, respectively, which is significantly positive and has a relatively large magnitude. This indicates that suppliers in the eastern coastal areas, benefiting from sound digital infrastructure, the dense agglomeration of innovation factors, and an open market environment, have lower digital access costs and richer knowledge absorption channels. The green innovation empowerment effect caused by downstream customers’ digital transformation is stronger, and they are more likely to realize secondary knowledge spillover through intra-regional industrial clusters, forming a virtuous cycle of “empowerment–diffusion”.

In contrast, for the central region (columns (2) and (5)), the coefficient of DT on GI_C is 0.009, and on GI_M it is 0.023, which are both smaller in magnitude compared to the eastern region. For the western region (columns (3) and (6)), although the coefficient of DT on GI_C is 0.168 and on GI_M it is 0.260, considering the overall model fit and the potential limitations in digital infrastructure and high-end technical talent, suppliers in central and western regions are limited by weak digital infrastructure and a shortage of high-end technical talents. They may face the attenuation of the empowerment effect caused by the “digital divide” and need to rely on government-led digital assistance policies and targeted technical support from chain-leading enterprises, with their green innovation persistence being more dependent on digital knowledge.

Overall, these results confirm that the stronger impact of customer digital transformation on the motivation pathway than the capability pathway is not universal but is moderated by supplier ownership attributes, customer market power, and regional development conditions, with external resource dependence, short-term customer pressure, and market environment constraints being the core driving factors for the pathway gap.

5.6. Mechanism Test

As mentioned above, digital transformation may affect an enterprise’s sustainable capacity and driving force for green innovation through multiple paths. This section, based on the findings of Dai et al. [66], is used to construct the following model to test the above-mentioned mechanism of action. Among them, Med functions as the mediator variable, with other variables’ specifications fully aligned with the original regression model (1). ${\mathsf{\beta }}_1$ and ${\mathsf{\tau }}_1$ are the objects of concern in this paper. Only when both are significant can it be indicated that the mechanism is established.

$${\mathrm{M}\mathrm{e}\mathrm{d}}_{\mathrm{i},\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{D}\mathrm{T}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\theta }}_{\mathbf{n}}{\mathbf{X}}_{\mathbf{i},\mathbf{t}}^{\mathbf{n}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(2)

$${\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{t}}={\mathsf{\tau }}_0+{\mathsf{\tau }}_1{\mathrm{M}\mathrm{e}\mathrm{d}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\tau }}_2{\mathrm{D}\mathrm{T}}_{\mathrm{i}.\mathrm{t}}+{\mathsf{\theta }}_{\mathbf{n}}{\mathbf{X}}_{\mathbf{i},\mathbf{t}}^{\mathbf{n}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(3)

5.6.1. Mechanism Testing of Green Innovation Sustainability Capability

Firstly, inspired by the approaches of Yi et al. [67] and Brav et al. [68], we employ the count of collaborative patent filings (NPA) in critical industries as a proxy for the knowledge spillover effect. Professional joint applications in key industries are a direct manifestation of knowledge sharing and collaboration among different entities within specific industries to achieve technological breakthroughs or the output of innovative achievements. Through a quantitative examination of collaborative patents in the sector, the scale and frequency of the knowledge spillover within the industry can be quantified more intuitively and accurately, providing a reliable quantitative basis for studying the degree, trend, and impact of the knowledge spillover at the industry level, effectively reflecting the level of the knowledge diffusion outward through joint collaboration within the travel industry. Secondly, when the efficiency of the supply and demand coordination is low, each link in the supply chain lacks an effective information exchange and joint decision-making mechanism. The real demands of downstream customers are difficult to accurately convey to the upstream. Each entity often makes predictions and decisions based on its own local information, resulting in the distortion of demand signals layer by layer and the intensification of the bullwhip effect. This further leads to problems such as production plan chaos, inventory overstock, or shortages among upstream suppliers. Conversely, if the efficiency of the supply and demand coordination is enhanced—and each link can grasp the real demand from a global perspective, reducing the blindness of the individual decision-making—the degree of the distortion of demand signals will be significantly reduced, and the bullwhip effect will be alleviated. The efficiency of supply and demand coordination is directly reflected in the bullwhip effect. Therefore, this paper adopts the bullwhip effect as its measurement index. Building upon the method of Shan et al. [69], we use the ratio of the enterprise production and demand fluctuations as the proxy variable for the bullwhip effect, as shown below:

$${\mathrm{B}\mathrm{u}\mathrm{l}\mathrm{l}\mathrm{w}\mathrm{h}\mathrm{i}\mathrm{p}}_{\mathrm{i},\mathrm{t}}={\displaystyle \frac{\mathsf{\sigma }\left({\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{d}\mathrm{u}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}_{\mathrm{i},\mathrm{t}}\right)}{\mathsf{\sigma }\left({\mathrm{D}\mathrm{e}\mathrm{m}\mathrm{a}\mathrm{n}\mathrm{d}}_{\mathrm{i},\mathrm{t}}\right)}}={\displaystyle \frac{\mathsf{\sigma }\left({\mathrm{C}\mathrm{o}\mathrm{s}\mathrm{t}}_{\mathrm{i},\mathrm{t}}+{\mathrm{I}\mathrm{n}\mathrm{v}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{y}}_{\mathrm{i},\mathrm{t}}-{\mathrm{I}\mathrm{n}\mathrm{v}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{y}}_{\mathrm{i},\mathrm{t}-1}\right)}{\mathsf{\sigma }\left({\mathrm{C}\mathrm{o}\mathrm{s}\mathrm{t}}_{\mathrm{i},\mathrm{t}}\right)}}$$

(4)

Among them, σ (∙) represents the standard deviation of the variable, Production represents the production volume of the enterprise, Inventory represents the net inventory value of the enterprise, and Cost is the main business cost of the enterprise and serves as a proxy variable for the Demand of the enterprise.

The bullwhip effect, as a classic phenomenon of distorted and amplified demand signals in supply chains, is typically measured through indirect indicators. While this alternative approach reflects certain characteristics of supply–demand mismatches, it struggles to directly capture the core efficiency level of supply–demand coordination. The Inventory Turnover Ratio (ITR), as a core metric for measuring inventory monetization efficiency, can more accurately capture the dynamic equilibrium state of supply–demand coordination. When the coordination efficiency improves, demand signals transmit more smoothly, and inventory management becomes more precise. Upstream suppliers can achieve on-demand stocking, leading to a steady optimization of the ITR. Conversely, insufficient coordination often results in abnormal fluctuations in the ITR. Therefore, supplementing existing analyses with the inventory turnover rate (calculated here as cost of goods sold divided by ending inventory balance) provides a tangible representation of supply–demand coordination effectiveness from an inventory operational efficiency perspective. This makes it a crucial complementary metric for evaluating supply–demand coordination efficiency.

Table 13. Mechanism test on the sustainable capability of green innovation.

Variable	NPA	Bullwhip	ITR	GI_C	GI_C	GI_C
	(1)	(2)	(3)	(4)	(5)	(6)
DT	0.232 ***	−0.169 **	0.692 **	0.071 ***	0.174 ***	0.027 ***
	(0.579)	(0.077)	(0.263)	(0.010)	(0.013)	(0.010)
NPA				0.036 ***
				(0.138)
Bullwhip					−0.186 ***
					(0.028)
ITR						0.071 ***
						(0.016)
Control	Yes	Yes	Yes	Yes	Yes	Yes
Id FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
N	2002	1553	2002	2002	1553	2002
R-sq	0.213	0.076	0.035	0.041	0.039	0.097

Note: Statistical significance is denoted as *** (p < 0.01), ** (p < 0.05), and robust deviations are reported within square brackets. The data for the mediating variables are sourced from CSMAR, CNRDS, and Wind databases.

validates the mediating roles of knowledge diffusion and supply–demand synergy in sustaining green innovation. By comparing the coefficients and model fits, the knowledge diffusion pathway and supply–demand synergy are complementary in driving green innovation sustainability. The ITR mechanism, as seen in column (6) with a relatively higher impact and a higher R-squared compared to columns (4) and (5), seems to play a more dominant role under the influence of digital transformation in facilitating the virtuous cycle for innovation persistence in this context. Specifically, the mitigation of information asymmetry and the improvement of inventory turnover, coupled with cross-organizational learning opportunities, collectively create this cycle, and among them, inventory turnover enhancement appears more pivotal. Thus, hypotheses H2a and H2b were verified, with the supply chain stabilization mechanism, particularly inventory turnover, showing a more prominent role alongside the complementary knowledge diffusion mechanism.

5.6.2. Mechanism Testing of Green Innovation Sustainability Motivation

Firstly, drawing on Jin’s approach [70], the Herfindahl Index (HHI) is selected to measure industry competitiveness. This index is calculated by weighting the market shares of all enterprises within the industry. The framework incorporates the market concentration of industry leaders while accounting for competitive dynamics among niche players, accurately reflecting the market concentration of enterprises within the industry and thereby demonstrating the competitive landscape. This concentration ratio ranges from 0 to 1, where values approaching 0 reflect perfect competition with fragmented firm sizes, while values tending toward 1 indicate a monopolistic dominance with suppressed competitive intensity. This characteristic enables it to visually and comprehensively depict the degree of the concentration of industry competition. Secondly, the intensity of tax incentives is measured as the ratio of refunded tax amounts to aggregate fiscal obligations, computed as the ratio between refund amounts and aggregate tax payments (refunds plus taxes paid) in the cash flow statements, following the methodology established in He et al. [71].

The regression results in

Table 14. Mechanism test on the sustainable motivation of green innovation.

Variable	HHI	Tax	GI_M	GI_M
	(1)	(2)	(3)	(4)
DT	0.094 **	0.590 ***	0.126 ***	0.138 ***
	(0.013)	(0.241)	(0.014)	(0.015)
HHI			0.261 ***
			(0.074)
Tax				0.052 ***
				(0.016)
Control	Yes	Yes	Yes	Yes
Id FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
N	2002	2002	2002	2002
R-sq	0.0249	0.101	0.041	0.191

Note: Statistical significance is denoted as *** (p < 0.01), ** (p < 0.05), and robust deviations are reported within square brackets. The data for the mediating variables are sourced from CSMAR, CNRDS, and Wind databases.

show that industry competitiveness (HHI) and tax incentives (Tax) exert partial mediating effects in the process of DT influencing GI_D, thus verifying H3a and H3b. From the regression results, in column (3), when both DT and HHI are included, DT still has a significant positive effect on GI_D, and HHI also shows a significant positive impact. In column (4), with DT and Tax included, DT remains significantly positive, and Tax also has a significant positive effect. By comparing the R-squared values, column (4) has a higher R-squared than column (3) (0.039), indicating that the tax incentive mechanism explains more variance in the green innovation accumulation. Regarding the interaction between these mechanisms, both the industry competitiveness mechanism and the tax incentive mechanism play a role in facilitating green innovation accumulation and are to a certain extent complementary. However, the tax incentive pathway accounts for more of the variance in green innovation accumulation, suggesting it may be more dominant in enabling digital transformation to drive green innovation accumulation in this context.

5.7. Analysis of Regulatory Effects

Furthermore, we examine whether the supply chain concentration will affect the extent to which digitalization in the downstream boosts green innovation in the upstream and conduct an interaction effect analysis from the perspectives of the supplier and customer concentration. Supply chain focalization represents both structural dominance and network interdependencies [72]. As the supply chain concentration increases, cooperation among enterprises becomes more stable. A stable cooperative relationship can not only promote the spillover of information and technology among enterprises but also foster an enabling green regulatory climate for enterprises. Therefore, the supply chain density significantly alters the effect size of corporate digitalization on sustainable innovation outcomes. To empirically examine this moderation mechanism, we introduce an interaction term (Mod) in the model specification, while maintaining consistency in variable definitions with the benchmark regression framework.

$${\mathrm{G}\mathrm{I}}_{\mathrm{i},\mathrm{t}}={\phi }_0+{\phi }_1{\mathrm{D}\mathrm{T}}_{\mathrm{i},\mathrm{t}}+{\phi }_2{\mathrm{D}\mathrm{T}}_{\mathrm{i},\mathrm{t}}\times {\mathrm{M}\mathrm{o}\mathrm{d}}_{\mathrm{i}.\mathrm{t}}+{\phi }_3{\mathrm{M}\mathrm{o}\mathrm{d}}_{\mathrm{i}.\mathrm{t}}+{\mathsf{\theta }}_{\mathbf{n}}{\mathbf{X}}_{\mathbf{i},\mathbf{t}}^{\mathbf{n}}+{{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{t}}+\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(5)

5.7.1. Supplier Concentration

The supplier concentration (SC) reflects the degree to which suppliers rely on a few customers [73]. This article measures the concentration of suppliers by calculating the annual procurement proportion attributable to the top five vendors [74].

Table 15. The moderating effect of supply chain concentration.

Variable	GI_C	GI_M	GI_C	GI_M
	(1)	(2)	(3)	(4)
DT	0.069 ***	0.120 **	0.029 ***	0.154 *
	(0.011)	(0.011)	(0.079)	(0.101)
DT × SC	0.080 ***	0.202 **
	(0.124)	(0.111)
DT × CC			0.027 *	0.007 *
			(0.026)	(0.019)
Control	Yes	Yes	Yes	Yes
Id FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
N	2002	2002	2002	2002
R-sq	0.041	0.191	0.039	0.191

Note: Statistical significance is denoted as *** (p < 0.01), ** (p < 0.05), and * (p < 0.1); robust deviations are reported within square brackets. The data for the adjustment variables are sourced from CSMAR, CNRDS, and Wind databases.

reports the regression outcomes of the interaction effect. As evidenced in columns (1)–(2), the digital transformation–supplier concentration interaction (DT × SC) exhibits a statistically positive association at conventional levels, indicating that the supplier concentration positively moderates DT’s effect on the sustainability of green innovation. Under a high concentration of suppliers, the cooperation between enterprises and core suppliers is more stable and in-depth. Digital transformation can more efficiently drive suppliers to continuously improve and upgrade green technologies, processes, etc., facilitating the progressive development of green innovation and gradually advancing from basic green production to high-end green manufacturing. Meanwhile, the centralized supplier resources enable enterprises to more easily integrate the green resources and capabilities of suppliers under the empowerment of digitalization and reserve key elements such as technology, talent, and funds for their own green innovation, such as accumulating green patents and cultivating green R&D teams, to enhance the subsequent development potential of green innovation. Thus, H4a is verified.

5.7.2. Customer Concentration

The customer concentration (CC) denotes the level of dispersion of a customer’s own suppliers. If the customer concentration is low, the requirements for the green innovation of a certain supplier in its digital transformation may be relatively weak. If the customer concentration is high, customers will be more motivated to drive suppliers to continuously carry out green innovation through digital means to uphold circularity within supply chains. At this time, the sustainability of suppliers’ green innovation may be more easily stimulated. The client focus index is computed by normalizing major accounts’ sales, which is achieved by dividing the total sales of the top five accounts demonstrating peak annual expenditures [75]. Columns (3) and (4) in Table 15 indicate that the customer concentration moderates the relationship between customer digital transformation and green sustainability capabilities and motivation, with all corresponding moderation coefficients being significantly positive. In scenarios with a high customer concentration, enterprises are more driven by the green demands of key customers. Digital transformation can precisely align with changes in customer needs, prompting enterprises to continuously optimize the green innovation value chain to escalate sustainable innovation stages. On the other hand, when the customer concentration is high, for the purpose of satisfying the long-term green demands of core customers, enterprises will leverage digital transformation to lay out green innovation resources in advance and conduct forward-looking sustainable energy R&D and technological stockpiling to elevate their green innovation storage capacity and ensure its continuous development. This proves H4b.

Figure 3 synthesizes empirical evidence from the mediation analysis, confirming the hypothesized transmission pathways and visually presenting the effect directions and coefficient magnitudes of each pathway.

6. Conclusions and Implications

6.1. Discussion

This study methodically analyzes the influence of downstream enterprises’ digital transformation on the sustainability of upstream enterprises’ green innovation from a supply chain perspective, along with its underlying mechanisms. The empirical analysis yields a series of key findings. This section synthesizes the research discoveries and elucidates their conceptual contributions and managerial applications, while delineating investigational directions and acknowledging existing limitations.

(1)

Summary of Key Results

The empirical evidence confirms that downstream firms’ digital permeation generates transboundary impacts, which systematically enhance upstream green innovation sustainability. This manifests specifically in enhanced green innovation sustainability and strengthened driving forces, with the latter exhibiting a greater impact intensity. This conclusion remains robust after the stability testing and endogeneity treatment. Mechanism tests further reveal differentiated pathways between the two effects: the enhanced sustainability capacity primarily stems from knowledge spillovers and improved supply–demand coordination efficiency, while strengthened motivation relies on intensified market competition and heightened sensitivity to tax incentives. The interaction effect analysis indicates that the structural empowerment from the upstream market density and downstream market consolidation simultaneously enhances the sustainability, as evidenced in Chinese clusters.

(2)

Comparison and Extension of Current Literature

Current academic studies largely focus their attention on the short-term outputs of green innovation and the influence of individual actors, such as concentrating on short-term indicators like the quantity of green patent rights or examining the role of a company’s own digital transformation and unidirectional supply chain interactions. However, they often overlook the sustainability of innovation and multi-tier ripple effects across supply networks. This paper centers on the sustainability of green innovation, examining not only whether innovative actions occur but also focusing on the long-term persistence of innovation capabilities and the continuous supply of driving forces. It simultaneously reveals how downstream digital transformation exerts a sustained influence on upstream entities through dual pathways, while also considering supply chain spillover effects and refining their differentiated mechanisms. This approach addresses existing research gaps in dynamic evolution and cross-actor interactions, thereby enriching the theoretical framework.

(3)

Theoretical Contributions

By constructing a dual-pathway analytical framework, this study deconstructs the sustainability of green innovation into two dimensions, sustaining capability and sustaining motivation, identifying differentiated transmission mechanisms and offering a new analytical perspective for conceptual investigations of the dynamic evolution of innovation. Moving beyond the traditional single-firm analytical paradigm, this research centers on the spillover effects of downstream digitalization on upstream entities. It reveals the intrinsic logic of cross-firm green innovation coordination within supply chains, providing fresh empirical evidence for interdisciplinary research where supply chain management intersects with sustainability. Furthermore, by analyzing the differentiated moderating role of the supplier and customer concentration, it elucidates how supply chain structural characteristics influence the spillover efficiency of digital transformation, broadening research into the contextual mechanisms of interorganizational relationships in shaping innovation effects.

(4)

Practical Implications

For managers, upstream enterprises should proactively capture spillover benefits from downstream digital transformation: on one hand, by participating in downstream digital collaboration to absorb knowledge spillovers and elevate innovation capabilities for green sustainability; on the other hand, by leveraging market competitive pressures and policy incentive windows to strengthen long-term investment momentum in green innovation. For policymakers, efforts should focus on promoting supply chain digital collaboration policies; simultaneously, differentiated incentive policies can be designed for highly concentrated supply chains.

(5)

Research Limitations and Future Directions

The constraints associated with this research are primarily reflected in three aspects: First, the enterprise-level analysis concentrates on manufacturing activities within China, which could potentially restrict the applicability of conclusions to the service sector or multinational supply chains. Second, it fails to carry out a comprehensive examination of the differentiated spillover effects of various digital technologies. Third, the measurement dimensions of tax incentive measures remain insufficiently comprehensive, with existing indicators struggling to fully capture implicit subsidies or non-institutionalized support from local governments. Recognizing this limitation facilitates a more nuanced and objective interpretation of research findings. Future research may expand in two directions: First, conducting international comparative studies to investigate how structural environment differences influence the spillover effects of supply chain digitalization. Second, integrating carbon neutrality objectives to investigate the interactive effects between supply chain carbon footprint constraints and digital transformation, thereby providing more precise theoretical support for green supply chain governance.

6.2. Conclusions

The sustainability of corporate green innovation holds dual significance, extending beyond sustained capability development to encompass sustained motivation propagation throughout industrial ecosystems. This investigation adopts a supply chain perspective to examine how technological transformation influences enduring environmental innovation patterns. Through analyzing the cumulative and progressive characteristics of sustainable inventive processes, this study develops a dual-axis framework encompassing both the innovation maintenance capacity and momentum generation mechanisms, yielding the following key findings.

(1) Downstream digital diffusion exerts catalytic governance over upstream green innovation ecosystems, with measurable enhancements in both the sustainability preservation capacity and innovation momentum generation. The effect size differential confirms stronger digital institutionalization impacts on dynamic capability formation. These causal inferences withstand rigorous falsification testing, including instrumental variable approaches.

(2) The mechanistic pathways exhibit statistically divergent coefficients when transmitting downstream digitalization effects to distinct innovation sustainability dimensions: sustained capability is strengthened by promoting knowledge spillover and improving the efficiency of supply and demand coordination, and sustained motivation is enhanced by intensifying market competition and increasing tax incentives.

(3) The interaction effect analysis shows that the supplier and customer concentration can constructively regulate the mechanism of downstream digital transformation on the sustainability of upstream green innovation. The positive regulation of the supplier concentration is reflected in the accelerated implementation of the progressive effect and the promotion of the targeted investment of energy storage resources. The positive adjustment of the customer concentration is manifested in the rapid adjustment of the progressive rhythm and the continuous accumulation of energy to consolidate the stickiness of cooperation.

Table 16. Summary of hypothesis testing results.

Symbol	Hypothesis	Null	Test Method	p-Value	Conclusion
H1	Digital transformation can promote the sustainability of green innovation, which manifests as enhancing sustained capability and strengthening sustained motivation.	Digital transformation has no significant impact on the sustainability of green innovation.	Fixed Effects	0.000	Assumption Established
H2a	Digital transformation enhances sustained capability of green innovation by unleashing knowledge spillovers.	Knowledge spillover pathway is ineffective.	Mediation Effect	0.000	Assumption Established
H2b	Digital transformation enhances sustained capability of green innovation by improving the efficiency of supply–demand coordination.	The efficiency of supply–demand coordination pathway is ineffective.	Mediation Effect	0.000	Assumption Established
H3a	Digital transformation enhances sustained motivation of green innovation by intensifying market competition.	Market competition pathway is ineffective.	Mediation Effect	0.000	Assumption Established
H3b	Digital transformation enhances sustained motivation of green innovation by increasing fiscal incentive.	Fiscal incentive pathway is ineffective.	Mediation Effect	0.000	Assumption Established
H4a	Supplier concentration positively moderates the effect of digital transformation on green innovation sustainability.	Supplier concentration has no significant moderating effect.	Regulatory Effect	0.000	Assumption Established
H4b	Customer concentration positively moderates the effect of digital transformation on green innovation sustainability.	Customer concentration has no significant moderating effect.	Regulatory Effect	0.000	Assumption Established

summarizes the test results of all hypotheses, clearly presenting the content of each hypothesis, the corresponding null hypothesis, the adopted test methods, the significance level, and the final conclusion. All hypotheses reach statistical significance at the 1% level, comprehensively and intuitively evidencing the verification status of each hypothesis.

6.3. Policy Recommendations

This article reveals the intrinsic logic of how digitalization empowers supply chains to shift from “passive compliance” to “active sustainability”, directly responding to the core demands of sustainable supply chains with regard to resource efficiency, environmental friendliness, and collaborative governance. As evidenced by this investigation into digital sustainability enhancement mechanisms across extended green innovation, the present paper offers the following policy implications, aiming to assist enterprises at all links of the industrial chain to achieve green and sustainable development through digital collaboration.

(1) Embed blockchain-based reward mechanisms to accelerate intelligent transitions while securing circular innovation longevity.

First, establish a certification mechanism for “Green Digital Transformation Benchmark Enterprises.” Governments can set up dedicated support funds to prioritize funding for core enterprises in the downstream of industrial chains to undertake a “green + digital” integrated transformation, such as building industrial Internet platforms that integrate carbon footprint tracking and real-time energy consumption monitoring. Simultaneously, mandate benchmark enterprises to open supply chain data interfaces and establish unified standards for green data exchange within supply chains. This will reduce technical and cost barriers for upstream suppliers to access these platforms. Second, incorporate “supply chain green empowerment performance” as a mandatory metric in large enterprises’ ESG assessments. Require chain-leading enterprises with annual revenues exceeding RMB 10 billion to disclose quantifiable indicators in their ESG reports, such as the “number of green patents generated by upstream suppliers” and “percentage reduction in energy consumption per unit of supplier output.” Grant preferential procurement points to compliant enterprises, guiding them to achieve precise spillover effects through physical entities like “joint R&D laboratories” and “green technology training bases”, thereby disseminating digital knowledge and green technologies. Finally, promote the chain leader–supplier innovation alliance model. Select key industries like new energy and automotive manufacturing to support chain leader enterprises in establishing cross-entity innovation alliances. Implement intellectual property revenue-sharing mechanisms for knowledge exchange within these alliances to accelerate the commercialization of collaborative innovation outcomes.

(2) Smooth the transmission path of environmental technology leapfrogging.

For enhancing the sustainable capacity for green innovation, establish an “industry-level digital platform for green technologies.” For pollution-intensive sectors such as chemicals and textiles, the government will spearhead the creation of a digital platform integrating a “green technology repository, demand-matching platform, and expert consultation system.” This platform will consolidate green patented technologies from research institutes with environmental demand data from downstream enterprises. Through AI algorithms, it will achieve precise supply–demand matching, providing upstream suppliers with “customized green technology solutions.” To strengthen the sustained motivation behind green innovation, establish a “tiered tax incentive system for green patent output.” Refine the “carbon labeling and green procurement” linkage mechanism by mandating full-lifecycle carbon labeling on end products like home appliances and furniture. Incorporate the proportion of green-labeled products purchased into procurement evaluations for government- and state-owned enterprises. This approach leverages end market pressure to compel upstream suppliers to continuously invest in green innovation.

(3) Optimize the supply chain architecture to leverage the regulatory effect of the concentration.

On one hand, policies should encourage core downstream enterprises to establish long-term, stable strategic partnerships with key upstream suppliers, appropriately guiding the formation of a more deeply collaborative supplier structure. Supply chain alliances committed to green collaborative innovation may receive corresponding policy preferences. For core supplier groups, provide customized support for green transformation and digital capability enhancement, strengthening their capacity to effectively absorb downstream digital spillover effects and sustainably invest in green innovation resources. On the other hand, promote and support upstream–downstream enterprises in signing long-term contracts that include explicit green innovation targets, technical collaboration mechanisms, and cost/risk-sharing clauses. This provides stable institutional safeguards for sustained energy accumulation and deepened cooperation. Drive the establishment of supply chain data-sharing infrastructure based on trusted technologies like blockchains to reduce information asymmetry, enhance mutual trust in collaboration, and foster a stable cooperative environment for long-term resource investment in green innovation.

6.4. Limitations and Future Research

This article has the following deficiencies:

Firstly, we examined the impact of digital transformation on green innovation based on microdata at the enterprise level. In future research, we can focus on a specific industry for a more detailed exploration. Secondly, if the research scenarios are expanded to a cross-border context and the supply chain characteristics and institutional environments of different countries are included for comparative analysis, the global applicability and practical significance of the conclusions will be further enhanced. In addition, exploring and constructing a more scientific method for measuring the degree of the enterprise digital transformation is also a research direction worthy of in-depth exploration in the future.