High innovation in China’s new-energy-vehicle sector arises only when policy, market and normative pressures align — policy creates the core legitimacy boundary while multiple, functionally equivalent configuration paths (resource-driven, responsibility-buffering, or multidimensional pressure linkages) drive firm innovation across lifecycle stages.

Main Finding

Firm-level innovation in sustainable innovation ecosystems (illustrated with China’s new-energy-vehicle industry) is a configurational, evolutionary outcome: multiple institutional and environmental pressures must align to produce high-level innovation, and there are several functionally equivalent (equifinal) configuration paths. These configurations and their causal structures evolve across industry lifecycle stages and are organized in a core–periphery hierarchical topology in which policy-induced (coercive) pressure sets legitimacy boundaries while normative pressures guide technological direction.

Key Points

• Innovation is configurational, not additive: combinations of pressures and resources (not single factors) produce high innovation outcomes.
• Three empirically identified, functionally equivalent paths to high innovation:
• Responsibility buffering — likely reliance on normative/responsibility-related mechanisms to mitigate risks and enable innovation.
• Dual drive of policy and resources — strong interaction between policy (coercive) pressure and firm/industry resources.
• Multidimensional pressure linkage — simultaneous linkage of several pressures (e.g., policy, normative, market) producing synergy.
• Ecosystem evolution follows three stages:
• Regulatory dominance — early stage where policy legitimacy dominates.
• Market-imitation synergy — intermediate stage where market forces combine with prior pressures.
• Multipressure synergy — mature stage with coordinated multiple pressures driving advanced innovation.
• Pressures are unequal and hierarchically organized: policy-induced pressure is fundamental (core), normative pressure is more surface-level (periphery), and other pressures/resources occupy roles in-between.
• The study provides both identification of outcome-producing configurations and a deconstruction of the causal mechanisms that generate them.

Data & Methods

• Empirical context: China’s new-energy-vehicle (NEV) industry (applied case to reveal configurational and dynamic processes in an innovation ecosystem).
• Hybrid analytical framework:
  • Dynamic fuzzy-set Qualitative Comparative Analysis (fsQCA): used to identify temporally sensitive configurations of conditions (institutional pressures, environmental factors, resources) that correspond to high innovation outcomes; captures equifinality and path dependence across lifecycle stages.
  • DEMATEL–AISM (Decision-Making Trial and Evaluation Laboratory combined with Adversarial Interpretive Structural Modeling) using reverse deduction: used to deconstruct the causal structure among pressures found by fsQCA, map influence relations, and reveal core–periphery/topological hierarchies and mechanism pathways in the ecosystem.
• Contribution of method: combines configurational pattern discovery with causal-structure deconstruction to move from "which combinations work" to "how they causally interact and evolve."

Implications for AI Economics

• For empirical research on AI innovation ecosystems:
  • Move beyond single-variable regressions: adopt configurational methods (dynamic fsQCA) to capture equifinality and stage-dependent effects in AI sectors.
  • Use causal-structure tools (DEMATEL/AISM or similar) to unpack which pressures are leverage points versus surface-level guides.
  • Study lifecycle dynamics explicitly — policies that matter in nascent AI subfields may be different from those in mature applications.
• For policy design and industrial strategy in AI:
  • Policy is often the core lever (sets legitimacy and boundaries). Early-stage AI markets may require clear regulatory signals to shape safe and credible innovation paths.
  • Resource policies (R&D subsidies, talent pipelines, infrastructure) combined with regulatory clarity produce robust dual-drive paths to innovation — targeting both legitimacy and capability is critical.
  • Normative/social pressures (standards, professional norms, public trust) shape technological direction and adoption; they are not as foundational as policy but crucial for long-run diffusion and responsible AI.
  • Expect multiple viable paths to AI innovation — tailor interventions to firm capabilities and existing pressure configurations rather than one-size-fits-all policies.
• For firms operating in AI ecosystems:
  • Diagnose the current ecosystem stage and the configuration of pressures affecting you. Choose an innovation path consistent with your resource endowment and the dominant pressures (e.g., lean on normative legitimacy if resources are limited; seek policy-resource synergies if scalable capability exists).
  • Recognize that interventions targeting core (policy) levers can reconfigure the ecosystem more effectively than focusing only on peripheral norms or isolated resources.
• Methodological takeaway for AI economics: combining configurational approaches with causal-structure analysis helps explain not just correlation but mechanism and evolutionary dynamics — useful for designing policy interventions and firm strategies in rapidly changing AI domains.