As soon as a hurricane threatens to make landfall, reinsurerswant to know what their losses are going to be. Insurers need to startplanning their claims adjusting activities, and reinsurers want toknow how much they will likely pay out to their clients in reinsurance recoveries. 

One would think the catastrophe models are the logical places togo for this information. But the traditional models have notprovided much guidance to companies as events are unfolding. Evenafter an event, the traditional model vendors provide wide rangesof loss estimates that are not very useful to decisionmakers. 

Newer, open loss modeling platforms provide more timely andaccurate information on hurricanes in real time. This informationis valuable not only for real-time decisions but also as importantindicators of overall model accuracy.  

Related: Facing devastation head-on

Primary insurers

Primary insurers would like estimates of the numbers of claims,the average claim severity by wind speed band and location, andtotal losses. These numbers inform claims-handling decisions, suchas how many adjusters will be needed and where. 

Open loss-modeling platforms like RiskInsight provide thisvaluable information, which can be updated each time there is a newprojected track for the storm. Typically, the National HurricaneCenter (NHC) projects a hurricane's path and intensity fivedays into the future. This data is updated every six hours and isused by RiskInsight to generate a high-resolution wind intensityfootprint for the storm. This footprint is overlaid on an insurer'sproperty portfolio to assign a wind speed to each propertylocation.

Losses are estimated based on the wind speed, value of theproperty — including building, contents and time element coverages— and the model damage functions. This same process is repeated forstorm surge and inland flooding, if applicable. The image belowshows a schematic of the process.

The detailed claims and loss information becomes availableseveral days before landfall and is updated with each new advisoryfrom the NHC. After landfall, when the track and storm parametersare known, the final claims and loss estimates are derived. As theclaims come in for Hurricane Harvey and Hurricane Irma, insurers are findingthe open loss modeling platform estimates generated inreal time are matching their actual losses quite well.

Reinsurers

For reinsurers, dozens of their ceding company clients may beimpacted by a live hurricane event, and it may not be practical torun all of those insurer portfolios multiple times a day. Instead,reinsurers can rely on updated total industry loss estimates forhurricanes in real time. The industry loss estimates are generatedusing the same process as described previously, but the model inputis a high-resolution database of total insuredproperty values versus an individual insurer'sportfolio. 

Reinsurers using open loss modeling platforms will havepre-generated information on their losses and market shares bylandfall point. For example, they will know that for hurricanesstriking the Galveston-Houston area, they might have 1% of theindustry loss, and for hurricanes striking Miami, they'll have aloss that's about 2% of the industry number. 

So, as a hurricane is unfolding in real time, reinsurers can usethe projected landfall point and the industry loss to develop acredible estimate of what their losses will be with each new NHCadvisory. Reinsurers can also run the portfolios of the insurersmost likely to have the largest losses based on geographicalexposure concentrations, which will also be known to them inadvance. After landfall, reinsurers can run the final storm trackthrough their entire portfolio to get a more detailed and accuratebreakdown of their likely losses by ceding company. 

Sophisticated users of open loss modeling platforms canalso create alternative tracks using other meteorological sourcesand advanced tools built into the modeling platform. These featurescan be used to provide a range of loss estimates to decision-makersversus one number.

Why real-time loss estimates matter 

Real-time loss estimates provide important information todecision-makers as the data is received. But perhaps even moreimportant, actual events provide the only true tests of theaccuracy of a catastrophe model. 

The catastrophe models simulate thousands of potential futureevents and provide loss estimates for those hypothetical events.How can reinsurers have confidence in the loss estimates for thosehypothetical events if a model can't produce accurate lossestimates for real events?

In the past, catastrophe models have been evaluated based on theunderlying science.  But all of the major models are basedon the same science. The data on historical US hurricanes, forexample, is updated and maintained by the NHC and is in the publicdomain. The research and literature used for most of the modelformulations and assumptions are essentially the same acrossmodelers, as can be verified by reviewing the Florida Commissionon Hurricane Loss Projection Methodology submissions.

The differences between the models stem not from differentscience, but from how well the science is implemented and to theextent the implementation process leads to consistent and accurateloss estimates. Catastrophe models are not scientific tools —they're business tools with science inside. The purpose of acatastrophe model is to provide reliable and accurate informationfor important business decisions — not just in real time, but wellbefore events occur. 

New open catastrophe models are demonstrating more skill inproducing reliable and accurate numbers so reinsurers can havehigher confidence in all of the model output — for simulated andreal events.

Related: Catastrophe global re/insurance losses will exceed$100 billion in 2017, Fitch says

Karen Clark is president and chief executiveofficer of Boston-based catastrophe risk managementfirm Karen Clark & Co.