Automobile Usage-Based-Insurance: Improving Risk Assessment measured through telematics

Dissertation presented as the partial requirement for obtaining a Master's degree in Statistics and Information Management, specialization in Risk Analysis and ManagementAccurate risk estimation with proportionate fees is the cornerstone of insurance activity, a billion-dollar service industry. Due to progressive technological development, insurance companies are now able to improve their risk assessment in the underwriting process of automobile insurance. Through the installation of Onboard-diagnostic devices or with an application in the customers’ smartphones, insurance companies may measure behavioral and situational risk factors such as distance covered and driving habits. These new risk factors provide further information that helps the client’s risk evaluation beyond the traditional risk factors of customer and car specific. The objective of this research is to measure the increased prediction capacity of the claim predicting model by including driver behavior variables. A Generalized Linear model was applied, that includes not only the traditional risk factors, but also additional risk factors measured by telematics, and a new model-based ensemble predictor to a dataset with more than 3 million drivers. Results show that the incorporation of driver’s behavior variables increases the overall capacity of the model. By adding these behavioral risk factors, the actuarial accuracy is increased, leading to a more tailored approach of risk assessment and also awarding the clients that have safer conduct on the road and penalizing those with a more hazardous behavior prone to incurring in car accidents