Modeling compliance under multi-level governance: co-evolution of firms and regulators in China’s construction sector

Policy compliance in China’s construction industry is influenced by the interaction between central policy objectives, provincial regulatory implementation, and firm-level adaptation. Existing evolutionary game models have difficulty explaining why compliance trajectories differ across provinces and ownership types. This study develops a governance-coupled evolutionary game framework to examine the compliance strategies of state-owned enterprises and private contractors under multi-level governance. A governance coupling coefficient is constructed to measure the semantic alignment between central policy directives and provincial regulatory adaptations, with adjustments for provincial institutional capacity. Transformer-based policy-text embeddings are used to estimate policy similarity, and a hierarchical graph neural network is introduced to represent inter-provincial spillover channels that may affect local compliance payoffs. The resulting coupling and spillover terms are incorporated into a replicator dynamics model to simulate compliance trajectories under different regulatory regimes. The results suggest that stronger governance coupling is associated with higher equilibrium compliance and faster convergence, while ownership heterogeneity and inter-provincial spillovers generate differentiated regional pathways. The framework provides a modeling approach for evaluating differentiated regulatory instruments in construction governance, although further validation with firm-level and project-level data remains necessary.