Multiple Bonus–Malus Scale Models for Insureds of Different Sizes

How to consider the a priori risks in experience-rating models has been questioned actuarial community for long time. Classic past-claim-rating models, such as Buhlmann–Straub credibility model, normalize past experience of each insured before applying claim penalties. On other hand, classic Bonus–Malus Scales (BMS) generate same surcharges and discounts all insureds because transition rules within class system do not depend on risk. Despite quality prediction BMS this model could appear unfair many regulators it does recognize initial risk insured. In paper, we propose creation different BMSs type using recursive partitioning methods. We apply approach real data farm insurance product major Canadian company with widely varying sizes insureds. Because can change over time, study possible transitions between is also performed.