Insurers should use multi-model approach to reduce catastrophe model uncertainty

Insurers can reduce catastrophe model uncertainty and better estimate risk by adopting a multi-model approach, a Guy Carpenter report says.
In Managing Catastrophe Model Uncertainty: Issues and Challenges, Guy Carpenter examines three strategies for using multiple models to narrow the “uncertainty band” inherent with the use of cat models.
The uncertainty band is the plausible range above and below the cat model’s probable maximum loss (PML) estimate in which the true (but unknown) value may fall.
Advances within the modelling industry have reduced the width of the uncertainty band, Guy Carpenter says. But as insurers consider smaller areas of geography, additional uncertainties are introduced.
For example, on a national scale, Guy Carpenter estimates the confidence interval around a cat model’s lower value of range as 60% of the PML and 190% of the PML in the upper value range. On a localized level, the lower value of the uncertainty band would be 25% of the PML and 430% of the PML in the upper value range.
Multi-model techniques include “blending” (averaging) the model outputs, “morphing” one model’s output to reflect the characteristics of another model’s, or “fusing” model components (or at least outputs) into what is in effect a new model, the report says.
In each case, the correct selection of specific weight parameters and methodology is critically important. This selection needs to be informed by the adequacies and shortcomings of each model.
A multiple-model approach can also help smooth out the impact of individual model changes, which seem to have an increasingly acute effect on the industry.
“When a cat model says, ‘Your 100-year return period loss is $1,117,243,572,’ what it really means is that your 100-year return period loss is about a billion dollars, but it could be $600 million or maybe $2 billion dollars… or something like that,'” the report says.