Algorithm-Driven Delivery and Urban Traffic Externalities: Evidence from a Quasi-Natural Experiment in China

While algorithmic food delivery platforms optimize urban logistics, their physical negative externalities remain underinvestigated. Grounded in Prospect Theory, we argue that rigid algorithmic ETA mandates trap couriers in a psychological "loss domain," inadvertently incentivizing risk-seeking driving behaviors. Exploiting the staggered rollout of China’s dominant delivery platform, we employ a difference-in-differences (DID) design, augmented by Propensity Score Matching (PSM) and instrumental variables. Our empirical results reveal that platform entry significantly exacerbates urban traffic accidents. Furthermore, this detrimental effect is heterogeneously amplified by platform-side algorithmic friction, consumer-side digital impatience, and rider-side survival vulnerability. This study advances the Information Systems (IS) literature by unveiling the unintended physical consequences of algorithmic control, offering critical implications for responsible AI governance.