Sherry Thomas, Head of Catastrophe Management - Americas and James Burnett-Herkes, Senior Vice President
Could you afford to find that the portfolio you just acquired in North Carolina is more exposed to hurricane than previously assumed? What if next year’s Category 2 hurricane caused a loss in excess of 15 percent of your policyholders’ surplus? How will the changes in the U.S. Geological Survey National Seismic Hazard Maps impact your exposure to earthquake risk in the central and eastern United States?
These are questions of risk appetite and risk tolerance. The Institute of Risk Management defines risk appetite as “the amount and type of risk that an organization is willing to take in order to meet their strategic objectives.” Conversely, risk tolerance is often expressed as a quantitative measure of the maximum amount of each type of risk one is willing to take.
In the context of insurance and catastrophe modeling risk appetite is about how much risk you are willing to take and risk tolerance is about how wrong you can afford to be when volatile loss experience rears its ugly head. Insurance executives and risk managers use catastrophe models to help inform these decisions, but how good are the models and how well do they represent your specific risks?
Catastrophe models have evolved significantly since they were first introduced in the 1980s. They now cover most natural and some man-made perils in countries around the globe, both in developed insurance markets and emerging markets. The proliferation of data sources and improvement in risk-defining exposure data reporting and an understanding of the underlying science and physical processes has enabled the creation of more sophisticated and detailed models. Their ubiquitous use throughout the industry has led to a greater dependence on catastrophe models for risk selection and management decisions. It follows then that there is a greater need for everyone from analysts, underwriters and risk managers to members of the C-suite, to understand the models - both for what they do well and where improvements could be realized.
When catastrophe models first came into use in the insurance industry, the innovators of the models were the resounding voice of science in the quantification of insurance portfolio loss estimation. As the collective use and adoption of catastrophe models expanded and the understanding of models matured to also include multiple views of risk, it was a turning point for exploring why views are sometimes so very different. Our industry does not yet have a set of standard protocols for assessing the suitability of catastrophe models or their use in making these critical risk management decisions. Guy Carpenter has made important investments and advances in this direction, introducing Model Suitability Analysis (MSA)®, a comprehensive framework to help insurers understand scientific validation tests, with complete documentation to prepare company-specific views of catastrophe risk.
MSA consists of a set of standard tests and protocols that benchmark the models against independent reference data for hazard, event frequencies, damage functions, losses and historical experience. These datasets are created by independent and credible third-party research institutions that have expertise in the respective subjects. Rather than reinventing the wheel and developing models that already exist, the MSA approach evaluates the scientific underpinnings of existing models to establish confidence where warranted, and to identify areas of uncertainty. Guy Carpenter aggregates this information into our MSA Knowledge Base, and establishes standard protocols that are efficient to execute and test all models using the same standard procedure to achieve homogeneity and fairness in the process.
The scientific appraisal aspect of MSA makes use of data from scientific and engineering authorities, which are defensible, reliable and independent of the model vendors. Such datasets allow a robust, transparent and independent evaluation of model performance relative to a balanced and consistent baseline.
Such review includes treatment of uncertainty inherent in all perils such as earthquake, hurricane and especially the severe convective perils. We also account for physically plausible but unobserved events in the evaluation, often with insight from our academic and expert partners.
Through an array of MSA tests, we help clients quantify how well a model represents the physical hazard being analyzed and the aspects of the model that directly impact loss volatility. The standard for model testing and evaluation through the MSA framework leads risk takers to more confident risk-informed decisions.