“A few mobile homes and an Air Force base, how much can it be?” So scoffed one Lloyd’s underwriter, when Hurricane Andrew made its landfall, barrelling through the Floridian town of Homestead, 25 years ago, on Monday, August 24 1992.
Back then, underwriters looked at premium but not exposure. The insurance industry was quite content with its maximum foreseeable loss (MFL) formula, adding up premium and multiplying it by a factor for a particular region, to predict catastrophe losses. The numbers were easy to manage.
One person who thought differently from the experts in the Lloyd’s underwriting room was Karen Clark, the first pioneer in what would soon become the catastrophe risk modelling business. She set up Applied Insurance Research as her own firm, now known as AIR Worldwide, and owned by Verisk since 2002.
Nowadays she again runs her own entity, Karen Clark & Company (KCC), a catastrophe risk modelling and consulting firm focused on those re/insurance clients dissatisfied with the ‘black box’ off-the-shelf modelling software, which later became the norm.
Clark had created the first catastrophe models in the mid-1980s, seven years before Andrew struck. “To give Lloyd’s credit, there were some reinsurers looking at our model before Andrew,” she says. “They just didn’t believe the numbers, which were already indicating that the loss potential was much higher than what insurers were predicting.”
In 1983 Hurricane Alicia had caused a $1bn insurance loss. Then Hugo in 1989 produced a $4bn loss. Before Andrew hit, insurers thought that $7bn was the sector’s worst case scenario, Clark notes, a small fraction of the figures derived from Clark’s then-upstart approach, which has since become standard.
Andrew therefore represented a watershed for the re/insurance industry’s attempts to quantify potential cat losses, leading to the switch to using a probable maximum loss (PML) approach, pairing geographically grouped exposures with historical or hypothetical hurricane data.
“It took a while,” says Clark. “We said the losses for Andrew could exceed $13bn and nobody believed that. The first estimates gradually went up. It took six months for the industry to realise they were going to pay $13bn. And as a direct consequence to the industry, Andrew triggered today’s Bermuda reinsurance market, with the first wave of Bermuda start-up reinsurers, the so-called ‘class of ‘93’, including Renaissance Re, Mid Ocean and Tempest Re.”
Cat modelling became mandatory, notes Clark. “It became critical for insurance companies to send their exposures to insurers. It started out at the county level, then by the Zip code, and now everybody provides geocode, so we know the lat-and-long of the individual property. So it not only brought cat modelling to the forefront, but also the importance of exposure collection,” she adds.
Swiss Re recently estimated that a modern-day Hurricane Andrew would cost estimated $80-100bn, some $50-60bn of which would be insured. The area struck has seen 35% population growth in the intervening 25 years. Clark agrees with the shape of these numbers. “I think most modellers would agree on that $50bn insured number,” she notes.
However, while Andrew was a big hurricane, it was still far from a worst case storm. Way back in 1926, the Great Miami hurricane devastated the city’s South Beach front. Fortunately, when Andrew made its landfall in 1992, it struck 20 miles south of bustling Miami.
“The most interesting thing is that, while if Andrew happened today exactly as it did, the industry would cope well, but if it made its landfall 20 miles north of where it did, so that it would be a direct hit on Miami, the insured losses are going to be above $200bn,” says Clark.
Swiss Re came up with comparably large figures. Shifting Andrew’s track north, and with today’s insured property values on the ground, the reinsurer estimated Andrew could cause a record $300bn of destruction, up to $180bn of which would be covered by private re/insurance.
“The industry is not prepared for that,” says Clark. “What I think is interesting is that we’re locked into one methodology today, just as we were before Andrew. Before Andrew, everybody was looking at the MFL; they thought that was enough; but it wasn’t. Today, companies are looking at the PML – and that’s not enough either. At KCC we came out with the Characteristic Event (CE) methodology, which shows companies where they’re most likely to go over their PMLs. Obviously Miami would be a hotspot, and that would be a major shock to the industry,” she continues.
The accepted PML to which the industry is managing today is about $125bn, Clark notes. “If Andrew hit Miami today, those companies with exposure concentrations in the area would certainly go well above their PMLs,” she adds.