The Anatomy of a Loss Estimate: Superstorm Sandy

Editor's note: Hemant Shah is CEO and president of catastrophe modeler Risk Management Solutions.

Superstorm Sandy was an extraordinary weather event, particularly given the destruction produced by its large storm surge in relation to the wind damage. Two weeks following the event, RMS issued its first and only estimate: insured losses would fall between $20 billion and $25 billion. So what goes into producing loss estimates and what have we learned about the art of post-event loss estimation from previous events?

Balancing the need for reliability with timeliness, we issued our loss estimate two weeks after Sandy made landfall. Our policy, based on extensive client consultation and a number of years of hard-earned experience, has revealed there is little benefit in rushing to publish and publicize an industry-loss estimate. It can be unhelpful if clients are forced to react to a modeler's estimate before we are even able to support their own internal estimations. 

It is not just about the “number,” but rather the insight that can help responsibly inform our clients' judgment of their losses. Moreover, each storm poses unique challenges; reliable loss estimation takes time. It is certainly not helpful to issue an estimate only to continuously revise it, or to issue a range so wide that it fails to provide actionable information. 

Within hours of Sandy's landfall, it was clear this was not a textbook hurricane: an unusual track; a large, diffuse and transitioning windfield; and a catastrophic storm surge despite borderline winds were some of its characteristics. Several days passed before floodwaters had receded and power and transportation resumed. Five days after the storm made landfall, some jurisdictions were still pumping water from flooded areas, and power was still out to tens of thousands of households. Previous major storms have shown that lifeline and infrastructure-recovery time can ultimately have a powerful impact on the total insurance bill. Combined with other drivers of loss, such as contingent-business interruption, it clearly takes time to fully understand and absorb the impact of these major events. 

Putting State-of-the-Art Modeling to the Test

Superstorm Sandy's large and destructive storm surge was far more damaging than its wind impact. The storm's broad windfield rapidly diffused as it made its way inland and, technically, Sandy was no longer a hurricane at landfall (hence its popular naming as “Superstorm” Sandy). Yet, the storm surge was equivalent to a Category 2 hurricane along the New Jersey coastline. This was not the first time the U.S. had experienced such a devastating and apparently disproportional storm surge: in 2005, Hurricane Katrina was a Category 3 hurricane at landfall with a Category 5 surge.

We were able to gain early insights into the potential magnitude of Sandy's surge damage by using the latest state-of-the-art hydrodynamic storm-surge model technology, in real-time, with inputs of the storm's characteristics from days before landfall. This advancement in storm-surge modeling has only been made possible in recent years. 

Complex hurricanes – such as those with extremely large windfields or those that decay but remain intense prior to landfall — as well as complex coastlines like the Hudson estuary on the New York/New Jersey border, have historically challenged the industry from a modeling perspective. Old-style surge models significantly understated the hazard and related losses. Learning from similar past extreme weather events, we incorporated a new way of modeling these unique physical features into version 11 of the RMS U.S. Atlantic Hurricane Model. 

The model encompasses a completely new way of analyzing storm surge, using state-of-the-art dynamic modeling to accurately represent a disproportionately large storm-surge risk compared to wind impact. 

On the Ground Research

In addition to this modeling effort, we examined thousands of damage reports and constantly evaluated diverse feedback loops. In the immediate aftermath of the event, we deployed two teams of engineers on the ground to conduct detailed surveys of the impacts along hundreds of miles of coastline. Two weeks later, a third team of engineers was deployed to assess the recovery efforts. Meanwhile, our catastrophe modelers in London and California worked around the clock to analyze the incoming data, integrating it with the modeling insights to generate information that insurers and reinsurers could run against their portfolios of accumulated exposures.

Drilling Down to Increase Precision

To calculate a more precise estimate of the losses, we dynamically modeled the surge street-by-street, distinguishing the flood risk by property based on the elevation and proximity of each building to the water's edge. Complex interrelationships between extensive power outages, disruption from flooding, widespread coastal property damage and the closure of transportation systems provided additional insights, as did other factors driving potential post-event loss amplification. 

It takes significant time and effort to collect and assess this wealth of data to ultimately develop a credible loss estimate. So how robust was RMS' range? Four months after our estimate was released, our range remains credible. Although, as with all extreme weather events it will take some time for all the claims to be resolved, $20 billion – $25 billion is consistent with current reporting from re/insurers across the industry.  

While we are confident in our estimate, we recognize this is not a time for complacency. Every catastrophe provides crucial insights and opportunities to learn. Our estimate was accompanied by a technical whitepaper that not only made our assumptions transparent, but also provided a roadmap for further learning. As we mine the claims data, we will learn more and incorporate improvements into our models.  Ultimately, it is not about 'the number'. As with our modeling in general, our response to Superstorm Sandy reflects our commitment to transparency and resiliency, by empowering our clients to take control of their modeling environment and make decisions that are informed by models, not dependent on them.