<p>Moral hazard in auto insurance arises when greater coverage leads policyholders to drive less cautiously, increasing claim likelihood. While prior research has primarily examined the direct correlation between coverage and claims, limited attention has been paid to the underlying causal mechanisms, particularly the mediating role of driving behavior. This omission constrains insurers’ ability to design effective risk mitigation strategies. To address this gap, we propose a novel approach that integrates mediation analysis with the residual inclusion method, explicitly accounting for the endogeneity of coverage selection. Simulation results demonstrate that our method surpasses the existing approach in both identifying moral hazard and uncovering its causal pathways. We apply the method to China’s automobile insurance market, leveraging IoT devices that monitor driving behavior before and after policy acquisition. Empirical findings reveal no evidence of moral hazard. Instead, additional commercial coverage is associated with reduced driving risk compared to compulsory coverage.</p>