The Role of Monitoring Effect in Risk Classification: Evidence from Telematics Adoption.

This article investigates the impact of the monitoring effect on the accuracy of customer type classification using temporarily tracked behavioral data collected via Internet of Things (IoT) devices, specifically telematics, in the auto insurance context. Drawing on a year-long randomized field experiment conducted by an insurance company and a major Asian car rental firm, the study finds that drivers significantly alter their behavior—reducing harsh driving—when monitored, and that this monitoring effect is positively correlated with their inherent risk levels. This correlation complicates the use of observed behaviors during monitoring to accurately infer true risk, substantially increasing misclassification rates if unaccounted for. Additionally, the monitoring effect partially persists after device deactivation through habit formation for most drivers, though some exhibit a crowd-out effect leading to riskier post-monitoring behavior. The findings highlight challenges for firms relying on temporary behavioral tracking for personalization and suggest that pilot studies estimating the monitoring effect’s relationship with inherent behavior can improve classification accuracy beyond traditional profile-based methods.