Evolutionary Game Analysis of Construction Suspension Supervision under Persistent Disasters in the Greater Bay Area

Abstract

This study aims to provide a methodological contribution by demonstrating how to incorporate exogenous environmental temporal dynamics into evolutionary game models. Unlike existing adaptive Evolutionary Game Theory (EGT) models that primarily capture endogenous behavioral feedback (e.g., adjusting penalties based on violation rates), this framework mathematically formalizes how external environmental duration structurally alters the payoff matrix over time. Using the persistent typhoons in the Greater Bay Area (GBA) purely as a narrative device, this study examines how disaster duration alters the cost-benefit structure of suspension supervision and identifies conditions to prevent regulatory failure. We built a tripartite evolutionary game model among government regulatory departments, construction units, and third-party subjects. A disaster persistence index (D) was explicitly embedded into compliance costs, accident losses, and governance benefits. We also treated detection probability as a dynamic variable. Replicator dynamics and numerical simulations were used to analyze equilibrium stability and parameter sensitivity. The system reaches an ideal safety equilibrium (compliant suspension, diligent performance, strict supervision) when compliance returns and diligence benefits outweigh violation and collusion gains. However, as disaster duration (D) extends, accumulating compliance costs quickly erode compliance incentives. This shift can drive the system into a dangerous state of "violation and dereliction" despite strict supervision. Sensitivity analysis reveals that compliance costs have a critical threshold effect. While higher fines promote compliance, their marginal utility diminishes. Furthermore, excessive collusion rent induces third-party dereliction, and rising regulatory costs threaten supervision sustainability. The model theoretically supports a shift toward dynamic governance. While mechanisms like duration-linked penalties or targeted compensation align with the mathematical conditions for safety equilibrium, these remain theoretical pathways requiring empirical validation before practical application.