RMS Chief Risk Officer Robert Muir-Wood: Version 11.0 model submitted to Florida Commission on Hurricane Loss Projection Methodology
Mar 31, 2011
The following article was published in PropertyCasualty360 (National Underwriter) on March 30, 2011:
RMS Explains Its Hurricane Model Revision
March 30, 2011 By Robert Muir-Wood
NU Online News Service, March 30, 2:04 p.m. EDT
Risk Management Solutions (RMS) released a new hurricane model at the end of February. Chief Risk Officer Robert Muir-Wood writes in to offer a closer look at the model and the science behind it:
Main Drivers Of Change
Catastrophe models are unlike other models of insured loss, having a relatively small number of historical data points compared to, for example, highly predictive auto claims models. Thus, catastrophe modeling methodologies and data continue to advance. Enhancements in our version 11.0 RMS hurricane catastrophe model have principally come through the availability of more and better data on hurricane behavior pre- and post-landfall, due to increased instrumentation, (we now have 10 times as much detailed wind speed data available compared to eight years ago when we last updated our hurricane hazard model), together with ongoing increases in computing power, enabling more simulations and at higher resolution than previously.
The scientific community has known through experience that there are many factors at play in determining how quickly hurricanes weaken after landfall; however, data from the National Hurricane Center (NHC) has not been sufficiently detailed prior to 1990 to allow us to understand how hurricanes really decay. Instead of waiting for hundreds more hurricanes to make landfall, we employed the best available numerical weather prediction models and simulated, at very high resolution, hurricanes moving onshore to see how they behaved after landfall. Through this we have been able to show how storm weakening relates to the size and speed of the storms and varies over different types of terrain—from the Florida Everglades to the dry Texas plains.
Advancements in the model also include complex and catastrophic storm surges—like those experienced during Hurricane Katrina, when even after the hurricane had weakened to a Category 3, the storm surge at landfall in Mississippi was the height expected of a Category 5 storm. The model captures the way in which the storm surge that flooded New Orleans first came from the east and then a few hours later from the north. The latest improvements in major flood defenses are also included.
We have now analyzed $18 billion of detailed claims data from storms that have occurred over the past 20 years. RMS worked closely with engineering consultants to re-examine code enforcement and building practices throughout the U.S. hurricane states, including regions where there have been few recent hurricane landfalls.
The changes in the model mean that wind risk has generally increased for all hurricane states on an industrywide basis. However, changes to individual portfolios vary considerably, depending on the region and line of business.
No New States Added
We have also released a suite of new and updated hurricane models for the North Atlantic Basin, including the Caribbean, Canada, Mexico and Central America. While inland risk is increasing in existing hurricane states, after weighing all of the evidence, we decided not to add any new U.S. states to the model.
The Time To Update
We found that in a couple of areas in the model, new data showed the level of risk was higher than previously identified. We have a responsibility to incorporate new information into our models whether it increases or decreases risk estimates. We had communicated the direction of the expected changes for several months before releasing the new model.
Models And Insurance Rates
Catastrophe models do not set insurance rates—insurers and regulators set rates based on factors related to local competition, supply and demand, by the costs of other loss elements beyond catastrophe risks, and ultimately by the approval procedures of state insurance authorities. We use exactly the same models for calculating catastrophe loss costs as for estimating how much capital insurers need to hold to guarantee they can survive a major loss. In all that we do, we aim to provide objective and unbiased risk information.
Model Validation And Approval
At every stage in developing the new model, we explore how each component, all the way through to loss reconstruction, can be calibrated against real data and validated in its performance in reconstructing detailed hurricane losses.
The new RMS hurricane model was developed over a three-year period, involving some of the leading hurricane climatologists as well as the leading scientist modelers within RMS. Expert third parties have peer-reviewed this model.
Over the coming weeks and months, RMS representatives will be visiting with state regulatory and legislative officials to educate them on the new model and discuss the implications for risk estimates within each state.
In Florida, models must undergo a rigorous review process with the Florida Commission on Hurricane Loss Projection Methodology (FCHLPM) before they are permitted for use in the rate-making process. We submitted the version 11.0 RMS hurricane model to the FCHLPM. The FCHLPM will perform a final review and certification of models in late May or early June.
The FCHLPM has certified the RMS model every year on all standards.