We’ve all heard the complaints: “Why are your rates uncompetitive?” or sometimes, “Why is AIG priced low on this risk?” How to determine the right rate for a specific risk is one of the age-old insurance questions.
Not all insurance risks are the same—and proper segmentation is key to charging what we believe is truly the correct amount—but how do we get there? Generalized Linear Modeling, or GLM, is a loss cost model based on statistics that not only helps us predict future losses, but also helps us price accordingly. To determine the final model, the actuarial, modeling and product teams validate the data and see how predictable it might be on our book of business. Based on the model’s direction, we determine a price.
When we launched GLM for property in 2014, it was the first by-peril rating structure in the high net worth space. To complement the actual rate structure approved by regulators, we built a Quality Score framework. This framework, which includes both attritional and CAT losses, is based on the difference between charged/approved premium and technical/adequate premium. Differences between the two are driven by attributes that are not allowed in pricing, but can be used in underwriting and modeling. Additionally, we determine adequate CAT premium through RMS models (InsightCAT), and include expenses, commission, targeted profit margin as well as a factor for concentration of risk.
Actuary and product have also built a similar GLM structure for auto that will begin rolling out this year. This, in conjunction with our newly-launched Auto Quality Score, should put us in a much more competitive position for risks we want to write.
GLM is an ever-evolving pricing methodology. As our book evolves and we are able to use more data, both in-house as well as third party, we expect our models and pricing to evolve with it.