Predictive modeling in property/casualty insurance has been most widely used in pricing, underwriting and marketing personal insurance products such as personal auto and residential. Commercial lines, however, may be a bridge too far to cross with predictive models.
Executive SummaryIn Part 2 in a series of articles on the pitfalls of predictive modeling, Ira Robbin discusses the applications of models to P/C insurance—particularly credit scoring, telematics and claims predictions—as well as why models are more suited for personal lines than commercial. For the complete report, download the entire whitepaper, "Predictive Modeling Pitfalls Whitepaper: When to Be Cautious."
This three-part series focuses on the use of predictive models in property/casualty insurance and illustrates several pitfalls.Here, in part 2, we review the pricing and claims applications of predictive models in personal lines pricing, and examine some questions about the applicability to such models to commercial lines pricing.
Applications in those lines are well suited for predictive modeling, since there are a large number of policyholders and extensive, reliable information on their attributes. The number and size of the claims for each policyholder are known over many policy periods. There are enough claims and the losses are usually small enough that any real effects come through and are not overwhelmed by noise.
There is also a clear reward and a potentially large payoff: A company with a better model than its competitors might be able to find the most profitable niches, ones of which its competitors are not aware. It can also better avoid unprofitable classes.