Gone are the days when analysts simply fed data into their catastrophe models and accepted the results at face value.Increasingly, model users expect to understand what goes on under the hood, and want to adjust the model machinery to modify results.
With the release of Touchstone® V2.0, AIR is adding the first of several planned capabilities aimed at allowing users to further "own their risk." Touchstone's new Ground-Up Loss Modification feature allows the creation of customized views of risk by allowing users to modify event losses according to event parameters, region, peril, coverage, and line of business. Future releases will allow users to modify individual model components, such as frequency and severity as well as modifying underlying damage functions. Model users will find this first iteration-Ground-Up Loss Modification-invaluable for performing a number of functions, including:
- Accounting for non-modeled risk, such as tsunami following earthquake in regions where it is not yet a modeled peril;
- Examining different perspectives of risk (e.g., based on claims experience or internal research, or to perform sensitivity testing);
- Using loading factors to account for poor data quality or loss adjustment expenses.
To take one example, you might employ Ground-Up Loss Modification to modify modeled losses for all magnitude ≥8.0 earthquakes impacting counties along the western coast of the United States, raising them by a user-defined factor to account for tsunami loss.
Undoubtedly, the ability to modify losses makes models more transparent and enables yet another avenue for sensitivity testing. However, to the extent that model components/configurations have been calibrated to observed loss output for historical events,modifying individual components could lead to a model that produces losses that bear no resemblance to observed claims. Accordingly,when exploiting Touchstone's loss adjustment flexibility, users should keep in mind the importance of maintaining a scientifically credible, internally consistent view of risk.
As noted in my earlier blog post, "Increasing Transparency and Flexibility," model validation is paramount: all models should be judged by the validation of loss output as well as by validation of individual model components. AIR, of course, firmly stands by the baseline model results, and modified model results will always appear with the baseline results.
For additional information on catastrophe model flexibility, take a look at a recent article on the topic by AIR President and CEO Ming Lee: "Model Flexibility: What's the Right Amount?"