Pricing Man-made Catastrophes

The following article written for Insurance Day by Robert Muir-Wood, managing director of global risk modeling at RMS, was published in February 2002.

For a decade, the $40bn industry loss events for which U.S. insurers bought protection in the international reinsurance market would be a CAT 4 hurricane through either Miami or New York; or a magnitude 7 earthquake beneath Los Angeles. The collapse of the World Trade Centre towers on September 11th 2001 showed that the universe of risk to which reinsurers had been offering coverage was larger than they had previously recognised. As terrorism coverage evolves, what are the risks that reinsurers carry in their U.S. portfolios, and how should a reinsurer price ‘Man-Made CAT’ alongside the hurricane and earthquake risk with which they are more familiar?

Losses from the WTC are now expected across more than 22 different categories of insurance business, with realistic loss estimates between $35 and $70 billion. For natural hazard risk, many insurers manage their accumulations to between the 100 and 250 year loss level. As a point of reference, RMS places such losses between $50 and $70 billion. What characterises the WTC event is the different mix of losses to those in a typical natural catastrophe. Approximately 80% of the loss in the 1994 Northridge earthquake was property related. For the WTC disaster, that figure has declined to 30% with Business Interruption (25%) and injury related claims (35%) taking center stage.

As a property loss alone, the WTC event was not so exceptional – the size of loss expected about once a decade – with no U.S. CAT loss approaching this magnitude since 1994. However, it is in the other lines of business that the event has shocked the markets, with the return period for the Workers’ Compensation loss being effectively ‘off the scale’ of the actuarial models for simultaneous workplace accident and mortality. Also unprecedented was the degree to which losses correlated across separate silos of risk for Life, Business Interruption, Property, Aviation Hull, Workers’ Compensation etc. that had previously been considered independent.

Of course for reinsurers the most astonishing feature of the WTC events was that protection against this category of Man-Made super-CAT loss was effectively being offered for free. So one natural response is to see if there are other classes of super-CAT events out there. In Natural CAT these include unexpected large earthquakes beneath New York City, the largest potential caldera-collapse explosive volcanic eruptions, as well as asteroid impacts. However the risk of any of these can be reasonably estimated from available statistics, eturn periods lie beyond 1,000 years. Therefore it is among Man-Made incidents that the ‘invisible CAT risk’ is chiefly concentrated.

Should the insurance industry exclude terrorism coverage? The history of insurance exclusions is not encouraging. Fire following earthquake was excluded from ordinary fire policies following the 1906 San Francisco earthquake, flood coverage in the Netherlands after the 1953 flood, and terrorism reinsurance in London following the 1993 and 1994 IRA bombs. In each case the response was probably the correct one: to force political responsibility to mitigate or reduce the risk through direct actions, such as creating major new sea-defences along the Dutch coast, installing fire-protection water-mains or increasing homeland security. However, for any of these excluded perils the only occasion when the insurance industry might justifiably point to the wisdom of their decisions was after the fire following the Great Tokyo earthquake of 1923. Perversly, even though fire-following earthquake had been excluded, the Japanese government forced the insurers to pay up, as far as their reserves would allow. Having successfully forced the mitigative action, in each case the insurance industry would have been better to move back in again to take the risk, and the associated premium.

For most potential categories of terrorism incidents, portfolio risk management is easier than for other more distributed sources of loss, such as a hurricane. In several major U.S. cities, there is over $30bn in insured buildings and contents property value within a 1km radius of a downtown location. There are daunting complexities in tracking information across a modern urban matrix and of defining insured values for a portfolio of highly customised risks, brokered and underwritten through separate channels. The industry underestimated its exposures in and around the WTC by as much as 40%. RMS is already working with a number of insurers to develop tools that facilitate the collection and processing of exposure data in 3D at building resolution. In the next stage of the information foodchain, reinsurers need to demand to obtain the same high resolution spatial data by which to track accumulations across multiple cedants, aggregating their Workers Compensation and Property portfolios.

There are two classes of Urban CAT events: accidental and intentional. Accidental events occur on all scales – nine days after the collapse of the World Trade Center towers an explosion at a fertiliser factory on the edge of the City of Toulouse highlighted the potential for accidental Urban Catastrophes – with damage levels of 95% of property values at a distance of 1km from the site, and a total insurance bill of around $2Bn. Chemical release catastrophes, such as those of Bhopal or Seveso illustrate the potential for industrial accidents on a far larger scale and consequent insurance impact. Nuclear power plant incidents on the scale of Chernobyl would also cause massive business interruption claims, with lengthy arguments about who should bear the responsibility for the loss. The accidental/intention distinction is itself blurred: was the Channel Tunnel fire started deliberately? If the wind had been stronger on September 11th would the WTC fires have spread through the city blocks of the downtown financial district?

It is important to think beyond simple repeats of the “aircraft into high-rise” type “vertical” event scenario, where the loss is correlated with the spatial geography of the buildings, people, and infrastructure located within close proximity. Other events could be more “horizontal” in nature, affecting more dispersed targets or economic sectors, where the correlation occurs through networks or dependencies (i.e. biological, chemical or IT network disruptions).

Terrorism has an innate logic – it could be argued that terrorists are rational individuals, deliberate in their acts: knowing that the medium is the message. The impact of any terrorist operation is proportional to its cost of execution: cost in terms of planning time, number of suitable martyrs, logistical magnitude. A skeptic might think that only terrorists would be able to develop a satisfactory model, yet even terrorists are unaware of how likely they are to be thwarted in their activities. The problem is eminently stochastic, and little different to other human risk problems regularly tackled by risk analysts, involving the modeling of sabotage risk in a large industrial corporation or of a rogue trader in a bank’s compliance operations.

While individual event scenarios can provide a useful stopgap for some aggregation management, it is only fully probabilistic modeling that gives the ability to price and quantify risk transfer. Twenty years ago few in the insurance industry thought it was possible to model hurricane risk, to price the hurricane element of a homeowner’s insurance premium or determine the loss threshold and cost of a reinsurance treaty. Now there is little hurricane insurance sold or transferred to a reinsurer without being modeled, often several times through each transaction. As a reflection of the stochastic links between natural and man-made catastrophes: “kamikaze” is Japanese for the divine typhoon wind that protected the islands from the urban destruction that would inevitably have followed a Mongol invasion. With all its inherent uncertainties, work has already started in RMS to build models for Man-Made CAT so that insurers and reinsurers can price and manage the ‘invisible’ and unquantified risks that lurk in their portfolios.