07 November 2013

Insurance Risk North America: Insurers warn of over-reliance on 'captivating' models

Jacob Rosengarten, chief enterprise risk officer at XL Group, said insurers had to work hard to find the balance between using their models and relying too heavily on them. "These models are mathematically remarkable creations; they are so elegant and so impressive. The mathematics is captivating, but we've all been round the block a few times, we all know that sometimes models fail", he said.

"Over-reliance is a big deal. That said, if there are things out there that can be figured out by mathematical models and you don't do it, shame on you because you could have. On the other hand, if you run the model and believe that the model always has to be right, shame on you because you also know better. So it's this constant tension, it's an incredibly important tool, but common sense still has to play a major role here."

Insurers must have a robust vendor-selection process when choosing their models, delegates heard. Frank Cassandra, senior vice-president at MetLife, said insurers should have consistent standards for judging third-party models, especially if they have operations overseas. 

The vendor should be able to produce test results an insurer can examine before purchase in order to be comfortable that the insurer's products will be modelled as intended, Cassandra said. Insurers, said Cassandra, must also verify that their vendors have appropriate procedures for modifications and updates over time. "Models can go stale over time. When we implement a vendor model, we want to understand pretty clearly what the methodology's going to be for modifications and updates."

Models often come pre-set with certain key assumptions or a model may be seeded with certain values. Insurers had to take time to understand these and assess their applicability to their own business requirements, he said.

Full article (Risk.net subscription required)

© Risk.net