Evolutionary Game and Simulation Study on Collaborative Governance in Innovation Ecosystems: A Four-Party Analysis Involving Government, Core Enterprises, Supporting Enterprises, and S&T Service Institutions

Effective governance of innovation ecosystems depends on strategic coordination and dynamic adaptation among multiple heterogeneous actors. However, existing studies are yet to systematically unravel the endogenous evolutionary mechanisms underlying multi-agent strategic interactions. This paper constructs a four-player evolutionary game model encompassing local government, core enterprises, supporting enterprises, and S&T service institutions, and examines the dynamic evolutionary paths of collaborative governance in innovation ecosystems under the assumption of bounded rationality. The model integrates key mechanisms such as fiscal incentives, knowledge spillovers, free-riding benefits, and service innovation. By employing replicator dynamics equations, we analyze the evolutionary stability conditions of different strategy combinations, and we use numerical simulations to investigate the impact of core parameters on the system’s evolutionary trajectories. The findings reveal that: (1) core enterprises’ deep collaborative innovation plays a pivotal role in the system’s evolution, with their strategic choices being jointly influenced by government subsidies and default penalties; (2) when the subsidy intensity is too low, the system converges slowly; when it is too high, fiscal sustainability constraints and strategic subsidy dependence may undermine the system’s long-term stability—the optimal subsidy interval depends on the relative gap between the collaborative innovation benefits and costs of core enterprises; (3) supporting enterprises’ free-riding behavior exerts an endogenous erosion effect on system collaboration, the intensity of which is jointly moderated by the degree of knowledge spillover and the relative benefits of free-riding; and (4) convergence to a high-level collaborative equilibrium requires the simultaneous satisfaction of incentive compatibility conditions for all agents—insufficient incentives for any single agent may cause the system to slide into a low-level collaborative trap.