At the start of the 1992 hurricane season, catastrophe-modeling company AIR had been in businessfor five years, and its hurricane model was projecting that a majorhurricane striking Miami could cost the insurance industry morethan $30 billion.

This seemed like an outlandish figure at the time — few peoplebelieved this number or even knew what a catastrophe model was.

It was difficult to convince insurers and reinsurers that a newapproach to estimating catastrophe losses was needed.The reinsurers were quite sure their tried-and-true traditionalformulas were robust and they needed no other tools. After all,they had been making money for decades. Insurers simply relied onthe brokers to determine how much reinsurance they needed, andthat, too, seemed to be working fine.

Traditional approaches

The traditional approaches seemed adequate because no major hurricane had madelandfall near a highly populated area in decades. And while the1970s and 1980s were relatively inactive with respect tohurricanes, property values along the coast were growingexponentially. But insurance companies were not monitoring theirexposures in 1992 — insurers and reinsurers were using premiums toestimate potential catastrophe losses.

Hurricane Hugo, instead of serving as a strong warning, had madecompanies even more complacent. The insured loss from the 1989storm was only $4 billion, even though the hurricane was a Category4 storm that "hit" Charleston, S.C. Actually, Hugo's most intensewinds occurred well to the north of Charleston along a sparselypopulated area near McClellanville, S.C. So even after Hugo, theindustry thought the probable maximum loss for the market as awhole for a hurricane hitting anywhere along the coast was $7billion.

This water tower, shown Aug. 25, 1992, a landmark at FloridaCity, Fla., still stands over the ruins of the Florida coastalcommunity that was hit by the force of Hurricane Andrew. The stormdamage to the South Florida area was estimated at $15 billion,leaving about 50,000 homeless. (Photo: Ray Fairall/APPhoto)

Enter Hurricane Andrew

As Hurricane Andrew was making landfall on Aug. 24, 1992, therewas again disbelief when the first hurricane model projected thatthe loss could exceed $13 billion.

In fact, it would take more than six months for the industry tofully realize it was going to pay out $15 billion in losses. And itdidn't take much math to calculate that had Andrew hit farthernorth, closer to downtown Miami, the loss could have been $60billion.

Fast forward to 2016 and complacency has once again set in.

It's been 10 years since a hurricane has made landfall inFlorida — unprecedented in the historical record — and no majorhurricane has made landfall near a populated area in the UnitedStates since 2005's Katrina. Coastal property values in Florida andelsewhere have continued to swell. The catastrophe models have nowbecome the traditional tools, and the industry has become far toocomfortable relying on the model-generated probable maximumlosses. 

Over the past few years, new tools have come into the market,and these innovative scientific techniques clearly show that thetraditional catastrophe models don't provide a full picture of aninsurer's loss potential. In many cases, the traditionalmodel-generated probable maximum losses give a false sense ofsecurity — just like the traditional approaches used in1992. 

And as in the pre-Andrew days, there's inertia, and even someresistance, to new scientific methodologies and insights intocatastrophe loss potential. 

Figure 1. Source: Karen Clark & Co.

Using the characteristic eventmethodology

Among the new tools is the characteristic event methodology thatprovides insurers with information on what size losses differentreturn period events will cause.

CEOs, boards and other senior decision makers in the insuranceindustry often mistakenly refer to their 1-in-100-year probablemaximum lossess as the 100-year events. But themodel-generated probable maximum losses don't refer to any specificevents — they're simply points on the estimated loss distributions.And what is also often forgotten is the 100-year probable maximloss shows the loss amount for which there's a 1% chance ofexceedance. 

Related: Here'swhat 100 years' worth of hurricane data says about losses andcatastrophe reinsurance

The characteristic events clearly illustrate where insurers canhave losses from the 100-year events that exceed their 100-yearprobable maximum losses. This is important underwriting informationthe traditional models don't provide, and while the probablemaximum losses may be similar, the characteristic event profilescan be very different between insurers. For some companies, the gapbetween the 1-in-100-year probable maximum loss and the loss fromthe 1-in-100-year event is significant as shown in the twocontrasting CE profiles for Company A (Figure 1, above) and CompanyB (Figure 2, below).

Figure 2. Source: Karen Clark & Co.

Obviously, an insurer will be better informed with bothperspectives on its large loss potential — the 100-year loss andthe loss from the 100-year event. The characteristic events helpinsurers take underwriting actions in order to avoid unpleasantsurprises along with highlighting clearly where companies can growprofitably without adding to peak exposures.

Figure 3. Source: Karen Clark & Co.

Multiple approaches provide betterinformation

Multiple approaches and perspectives on risk lead to betterpricing, underwriting and risk management decisions. Using multiplemodels is not the same as using multiple approaches because all thetraditional catastrophe models use essentially the same fundamentalapproach. Savvy underwriters, CEOs and other insurance companyexecutives instinctively understand the value of the newercharacteristic event method. 

It's the catastrophe modelers who need the most convincing thistime around because many don't see the need for anything other thanthe traditional modeling perspective. There's been too muchreliance on (and resource spent on) this one approach. Ironically,some of the 1992-era skepticism of the models would serve theindustry well today. Will it take Andrew II to awaken thecatastrophe-modeling community from its currentcomplacency? 

A storm like Andrew — the one-in-100-year characteristic eventin South Florida — will make a direct hit on Miami this year or ina future year. When it does, the resulting loss will be greaterthan many companies' probable maximum losses, the industry losswill likely exceed $200 billion and certain insurers will befinancially impaired. Although many insurer CEOs will be surprisedif they aren't given the vital characteristic event information, nometeorologist will be surprised because a Category 5 hurricane isjust as likely to strike Miami as Homestead,Fla.  

Credible scientific approaches

In a state such as Texas, the traditional model-generatedprobable maximum loss is almost meaningless. The Texas coastline isrelatively sparsely populated except around Galveston and Houston.The characteristic events profile in Figure 3 (above) illustratesthat if the 100-year Texas hurricane hits most areas along thecoast, the losses will not be too significant. However, a landfallanywhere around Galveston or Houston will produce an insuredindustry loss exceeding $50 billion, and just the right landfallpoints could produce losses in excess of $150billion. 

The traditional catastrophe models generally agree that the100-year probable maximum loss for Texas is around $40 billion. Sothe largest loss from the 100-year event is four times the 100 yearprobable maximum loss. Clearly, Texas companies managing to their100-year probable maximum loss could easily be overly exposed tothe 100-year-event. 

Given the well-known limitations of the traditional models andthe uncertainty surrounding catastropohe risk, it's wise to embraceall credible scientific approaches to understanding and managinglarge loss potential. It does not serve the industry well to staylocked into one way of thinking about catastrophe losses. Is thisdéjà-vu, and could history repeat itself this hurricane season?

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

Related: 4 tips to limit the risk of a big catastrophegetting bigger