Does Real-Time Feedback Make You Try Less Hard?: A Study of Automotive Telematics

Abstract

Problem definition: Mobile and Internet-of-things (IoT) devices increasingly enable tracking of user behavior, and they often provide real-time or immediate feedback to consumers in an effort to improve their conduct. Growing adoption of such technologies leads to an important question: “Does immediate feedback provided to users improve their behavior?” We study immediate (close to real-time) feedback in the context of automotive telematics, which has been recognized as the most disruptive technology in the automotive insurance industry. Academic/Practical relevance: Numerous automotive telematics providers claim unsubstantiated benefits from immediate feedback, while we still barely understand the implications of such feedback on user behavior. Given that feedback’s effect sometimes is ambiguous, at the same time such feedback-providing devices’ usage is increasing, it is important to study immediate feedback and identify the effect it has on human behavior, especially in important applications, such as automotive. This understanding is important given that other attempts to make driving safer have led to unintended consequences in the past. Methodology: Using proprietary data on driving behavior, as measured by several parameters such as harsh braking, speeding, and steep acceleration, we investigate the impact of the driver’s decision to review immediate feedback on driving behavior. We use instrumental variable regression to estimate the effect. Results: Contrary to the claims from multiple telematics providers, we find that on average, users’ driving performance after they review detailed feedback is nearly 14.9% worse than that of users who do not review their detailed feedback. This impairment in performance translates to 6.9%, or a one-year reduction in inter-accident time. Our results suggest that this deterioration is associated with increased speeding. Strong negative feedback (e.g., a sharp deterioration in performance) exerts a positive effect on short-term performance but this only happens for very large drops in performance (3% of cases). Furthermore, we demonstrate that drivers just below the insurance incentive thresholds exert greater effort following immediate feedback. Managerial implications: Our results provide a key message to firms employing immediate feedback – specifically that such technology can yield unintended consequences. Furthermore, we show that drivers should receive only strong negative (but not positive) feedback to improve their performance. Finally, our results suggest that insurance incentives should be continuous, rather than a step function.