GTI is vital for sustainable growth and carbon neutrality, especially in emerging economies such as China. Regional disparities in GTI efficiency are evident in differences in patent activity, product commercialization, and waste management (Duan et al., 2022; Xu et al., 2023). Underlying these disparities are systemic issues such as fragmented policy implementation, limited cross-sectoral collaboration, and uneven public engagement (Takalo & Tooranloo, 2021). While environmental regulations (e.g., emissions standards, green subsidies) and market competition are established drivers of GTI (Du et al., 2021), the nonlinear dynamics of multi-stakeholder interactions within innovation ecosystems remain insufficiently understood.

Traditional studies often adopt a reductionist lens, focusing on the isolated impact of variables such as government policies or corporate R&D investments. However, GTI is inherently an interactive and co-evolutionary process, requiring active engagement among enterprises, universities, governments, and civil society (Wang et al., 2024). For example, universities provide foundational research and knowledge spillovers, while enterprises drive commercialization through market-oriented adaptation (Holroyd, 2020). Despite this interdependence, existing frameworks fail to adequately explain how these actors collectively configure pathways for green innovation amidst diverse resource and institutional constraints.

Innovation ecosystem theory offers a comprehensive framework for understanding GTI as a complex system. Rooted in collaborative value creation, it unites diverse stakeholders—governments, firms, academia, and users—to drive GTI (Wang et al., 2021; Liu & Kim, 2024). This reframes GTI as a dynamic interplay of resource allocation, knowledge exchange, and institutional alignment, rather than a linear input-output process.

The digital revolution has transformed innovation systems into Digital Innovation Ecosystems (DIEs), which leverage modular platforms, big data analytics, and adaptive governance to reshape innovation trajectories (Hein et al., 2020). Digital platforms enable real-time collaboration between SMEs and research institutions, while AI-driven tools optimize resource efficiency in green product development (Luo et al., 2023). However, how DIEs foster GTI through configurational synergies remains empirically underexplored, especially in developing economies. Most studies rely on static, cross-sectional analyses, overlooking the temporal dynamics and path dependence essential to ecosystem evolution (Zeng et al., 2024).

Despite this interdependence, existing frameworks often fail to explain how these actors collectively configure pathways for green innovation amidst diverse resource and institutional constraints. To address this gap, this study develops an integrative theoretical framework based on the digital innovation ecosystem perspective, proposing a configurational model for understanding regional GTI.

Drawing on the DIE theory (Munodawafa & Johl, 2022; Filiou et al., 2023), this study constructs an integrative framework to examine how regional GTI emerges from the dynamic interplay of three key dimensions, as depicted in Fig. 1. Grounded in the "Actors-Environment" logic central to ecosystem research (Autio & Thomas, 2022), the framework comprises three interrelated components— Innovation Elements, Innovation Subjects, and the Innovation Environment—which collectively shape GTI outcomes.

Fig. 1.

A theoretical framework for DIEs to Drive GTI.

Innovation Elements (DE) represent the foundational data infrastructure that supports the ecosystem.

Innovation Subjects include Enterprises (EAI), Users (UEI), and Higher Education Institutions (HEII), which act as the primary drivers of innovation and commercialization, thereby extending the "Triple Helix" model into a digital context.

Innovation Environment comprises the Digital Policy Environment (DPE) and the Digital Cultural Environment (DCE), which collectively frame and constrain actor activities.

This triadic structure enables an analysis of GTI not merely as a sum of individual factors, but as the outcome of complex, co-evolutionary interactions across these dimensions. The subsections below elaborate on each component, with Table 1 providing a detailed summary.

Data Elements (DE): As the backbone of the DIE, data analytics enable firms to decode environmental trends, optimize resources, and accelerate green R&D. Real-time IoT data on energy consumption informs iterative improvements in clean production technologies (Song et al., 2019). Advanced machine learning models further enhance predictive capabilities, enabling simulation of eco-efficiency scenarios and prioritization of low-carbon innovations (Munodawafa & Johl, 2022). Cloud-based platforms integrate heterogeneous datasets, fostering cross-industry collaboration in sustainability monitoring.

Enterprise Innovation (EAI): Enterprises are the primary actors translating academic breakthroughs into market-ready solutions. Collaborative platforms linking enterprises with universities enhance knowledge transfer, shortening the time-to-market for green technologies (Vasiliev et al., 2020). Open innovation ecosystems enable firms to co-develop patents with startups, leveraging agile methodologies to scale circular economy solutions. Corporate venture capital further bridges funding gaps for pilot projects, aligning R&D with regulatory standards.

User Engagement (UEI): Consumer demand for sustainability exerts a strong market pull, incentivizing eco-design. Digital channels (e.g., social media) amplify user feedback loops, enabling rapid product refinement (Xu et al., 2023). Gamified mobile apps encourage active consumer participation in carbon footprint tracking, fostering behavioral shifts toward green consumption (Testa et al., 2021). Crowdsourcing platforms further democratize innovation by empowering users to co-design modular products for repairability and reuse.

University Innovation (HEII): Universities are critical engines of innovation, providing foundational research and skilled talent. Metrics such as university R&D expenditures and digital resource investments reflect their capacity to support ecosystem-wide innovation (Zabolotniaia et al., 2020). Interdisciplinary labs accelerate technology commercialization, with spin-off companies translating academic patents into climate-neutral applications. University-led hackathons also cultivate entrepreneurial skills, aligning curricula with industry demands for green digital competencies (Bai, 2021).

Digital Policy Environment (DPE): Regulatory frameworks—such as digital tax incentives and green procurement—set the "rules of the game," shaping collaboration incentives (Kunkel & Matthess, 2020). Dynamic policy instruments like sandbox regimes allow blockchain-based carbon credit trading to be tested, balancing compliance with innovation agility (Taeihagh, 2021). Cross-border data governance agreements further harmonize standards, reducing barriers to transnational green tech diffusion.

Digital Cultural Environment (DCE): Societal values favoring sustainability and digital openness foster trust and cooperation, reducing innovation network transaction costs (Chae, 2019). Grassroots digital literacy campaigns empower communities to advocate for eco-transparency, pressuring firms to adopt ethical AI practices. Cultural narratives around "smart green cities" also galvanize public-private partnerships, embedding sustainability into urban digital twins (Yigitcanlar et al., 2021).

This framework proposes that GTI outcomes emerge from the synergy among these dimensions rather than the influence of isolated factors. Regions with strong DPE (policy support for digital R&D), high HEII (university-industry partnerships), and robust DE (data analytics infrastructure) are more likely to achieve breakthrough innovations. Conversely, a deficiency in a key components, such as weak UEI (consumer demand) or low DCE (public environmental awareness), can create bottlenecks, disrupting configurational stability and ultimately hindering high level of GTI performance. Using Panel-QCA, this study captures these dynamic interactions, addressing gaps in static, single-period analyses.

This integrative approach advances DIE theory in the following ways:

Reconceptualizing GTI as an emergent outcome of multi-stakeholder configurations. Unveiling the temporal dynamics of how digital policies, cultural shifts, and technological capabilities co-evolve. Providing actionable, context-specific insights for policymakers to tailor ecosystem strategies across heterogeneous regional contexts.

This framework provides the theoretical foundation to empirically test how DIEs enable or constrain GTI, thereby offering a conceptual roadmap for sustainable transitions in emerging economies.

In summary, our theoretical framework posits that a high level of GTI is not the result of isolated factors but rather an emergent outcome of the synergistic configurations of DIE dimensions. This configurational hypothesis cannot be adequately tested with conventional linear methods. Therefore, to empirically investigate these complex causal recipes and their dynamics, we employ Panel-Qualitative Comparative Analysis (Panel-QCA), a methodology detailed in the next section.

QCA integrates qualitative and quantitative approaches by focusing on multi-factor configurations rather than isolated variables, making it ideal for analyzing complex causal phenomena (Ragin & Strand, 2008). Traditional QCA methods, such as fuzzy-set, crisp-set, multi-set, handle diverse variables (e.g., continuous, binary, categorical) effectively. However, they are inherently static, overlooking longitudinal variations and case heterogeneity. This can overgeneralize conclusions when configurational patterns shift over time. To overcome these limitations, this study applies Panel-QCA to assess the configurational impact of digital innovation ecosystems (DIEs) on green technological innovation (GTI).

This method was selected for its advantages over longitudinal alternatives like fixed-effects (FE) or random-effects (RE) panel regression models. While FE/RE models estimate the 'net effects' of individual variables over time (e.g., the impact of university R&D on GTI), this study centers on how GTI emerges from the concurrent interplay of multiple factors. Panel-QCA captures causal complexity, equifinality, and conjunctural causation over time—insights that regression models cannot provide (Linyu et al., 2024). Incorporating concepts such as pooled-consistency (POCONS), between-consistency (BECONS), and within-consistency (WICONS), enable Panel-QCA to capture temporal and individual effects in configurational analysis (Garcia-Castro & Ariño, 2016).

This study analyzes a balanced panel dataset of 310 province-year observations (31 Chinese provinces over the period 2013–2022). Data were primarily derived from provincial statistical yearbooks, the China Statistical Yearbook, and the China Environmental Statistical Yearbook. Green patent data were sourced from the publicly available records of the China National Intellectual Property Administration (CNIPA) and consolidated using the green patent classification codes provided by the World Intellectual Property Organization (WIPO). Missing values were addressed using standard interpolation and imputation methods.

The first stage tests whether any single antecedent condition is necessary for the outcome. Following Pappas et al. (2020), a condition is deemed necessary if its consistency score exceeds 0.9. As shown in Table 3, the consistency scores for all six conditions fall below this threshold, indicating that none is necessary for achieving high-level GTI. This result reinforces the central premise of this study: no single condition alone ensures high-level GTI. This absence of necessary conditions provides a strong justification for the subsequent sufficiency analysis, which examines how combinations of conditions produce the outcome.

Under the Panel-QCA principles, a consistency adjustment distance (the difference between pooled and within-group consistency) above 0.1 signals significant temporal variation that warrants deeper investigation (Hossain et al., 2024). Our analysis reveals that four conditions—EAI, UEI, HEII, and DPE—exceed this threshold, indicating their unstable association with high-level GTI over time. As illustrated by the temporal plots in Fig. 2, the consistency scores for these conditions fluctuate significantly between 2013–2022. For instance, the consistency of UEI remained stable until 2018. reflecting changing impacts of user demand and participation on GTI, with its influence fluctuating over time. Similarly, the consistency of EAI declined after 2019, suggesting diminished returns from initial corporate investments in GTI amid intensifying competition. The consistency of DPE, while initially high, gradually declined after 2015, likely due to the diminishing marginal effects of early government policies.

Fig. 2.

Consistency changes for individual factors.

In summary, the necessary condition analysis reveals that no single DIE element functions as a "single critical factor" for achieving high-level GTI. No single condition or its absence meets the 0.9 consistency threshold, provides compelling evidence against simplistic, single-factor explanations. This finding strongly supports the core premise that regional GTI emerges from the synergistic interplay of multiple interdependent elements. This result underscores the need to move beyond linear, variable-centric logic toward the holistic, configurational approach employed in the subsequent sufficiency analysis.

Our finding that no single DIE element constitutes a necessary condition for high-level GTI prompts an examination of how these elements combine into configurations sufficient for the outcome. Following QCA standards, we set a consistency threshold of 0.8, indicating that a configuration is a reliable subset of the outcome (Ragin & Strand, 2008; Fiss, 2011). To reduce contradictory configurations, we apply a Proportional Reduction in Inconsistency (PRI) threshold of 0.7, a conservative choice for exploratory research with theoretical ambiguities (Du & Kim, 2021). Given the large sample size (N=310), we used a case frequency threshold of 3 to ensure the resulting configurations reflect recurring patterns rather than idiosyncratic cases (Pappas et al., 2020). Consistent with QCA protocol, we report the intermediate solution as our primary finding and use the parsimonious solution to distinguish between core and peripheral conditions. Table 4 presents the results of the sufficiency analysis, detailing configurations leading to high-level GTI (Solutions S1-S2) and its absence (Solutions S3-S6).

Table 4.

Configurational analysis results.

A dynamic analysis of the solutions reveals that the between-group consistency adjustment distances for all configurations are below the 0.1 threshold, indicating overall temporal stability. However, for the configurations leading to the absence of high-level GTI, the within-group consistency adjustment distance exceeds 0.1, suggesting that case-specific dynamics play a significant role in these particular pathways.

Our analysis identifies two distinct pathways to achieve high-level GTI:

Solution 1

(S1): Comprehensive Synergy among Innovation Actors.